On the characterization of the n numbers such that any group of n P order has a given property Logan Crew 5 1 0 2 Advisor: Professor Thomas Haines n a J 6 ] R G . h t a m [ 1 v 0 7 1 3 0 . 1 0 5 1 : v i X r a Honors Thesis in Mathematics University of Maryland, College Park Contents 1 Introduction 3 2 Background Information 3 2.1 Sylow’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2 Solvable Groups . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3 Hall π-subgroups . . . . . . . . . . . . . . . . . . . . . . . . . 6 3 Cyclic Numbers 9 4 Abelian Numbers 11 4.1 Nonabelian Groups of Order the Cube of a Prime . . . . . . . 11 4.2 Burnside’s Normal Complement Theorem . . . . . . . . . . . 11 4.3 Determining the Abelian Numbers . . . . . . . . . . . . . . . 15 5 Nilpotent Numbers 17 5.1 Properties of Nilpotent Groups . . . . . . . . . . . . . . . . . 17 5.2 Nilpotent Numbers . . . . . . . . . . . . . . . . . . . . . . . . 18 5.3 Applying Nilpotent Factorization . . . . . . . . . . . . . . . . 21 6 Ordered Sylow numbers 21 6.1 Ordered Sylow Towers . . . . . . . . . . . . . . . . . . . . . . 21 6.2 Preliminary Results . . . . . . . . . . . . . . . . . . . . . . . 24 6.3 Determining the Ordered Sylow numbers. . . . . . . . . . . . 32 7 Supersolvable Numbers 33 7.1 R’edei’s First-Order Non-Abelian Groups . . . . . . . . . . . 33 7.2 Preliminaries of Supersolvable Groups . . . . . . . . . . . . . 35 7.3 Describing the Criteria for Supersolvable Numbers . . . . . . 36 7.4 Verifying the Criteria for Supersolvable Numbers . . . . . . . 37 8 Conclusion 44 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww 1 Introduction One of the classical problems in group theory is determining the set of positiveintegersnsuchthateverygroupofordernhasaparticularproperty P,suchascyclicorabelian. WefirstpresenttheSylowtheoremsandtheidea of solvable groups, both of which will beinvaluable in our analysis. We then gather various solutions to this problem for cyclic, abelian, nilpotent, and supersolvable groups, as well as groups with ordered Sylow towers. There is also quite a bit of research on this problem for solvable groups, but this research is outside the scope of this paper. This work is an exposition of known results, but it is hoped that the reader will find useful the presentation in a single account of the various tools that have been used to solve this general problem. This article claims nooriginality, butismeantasasynthesisofrelatedknowledgeandresources. To simplify terminology, if a positive integer n satisfies that every group of order n has property P, we will call n a P number. For example, if the only group of order n is the cyclic group, we will call n a cyclic number, and similarly for abelian number, nilpotent number, ordered Sylow number, and supersolvable number. Some notation: • C denotes the cyclic group of order k. k • (a,b) denotes the gcd of a and b. • Z(G) denotes the center of the group G. ′ • G denotes the commutator subgroup of G. • |G : H| denotes the index of a subgroup H in a group G. 2 Background Information Before continuing to the main results, we establish some foundations. 2.1 Sylow’s Theorem The results here are adapted from [1, p. 139]. Definition 2.1.1. Let G be a finite group, and let p be a prime dividing its order. IftheorderofGmaybewrittenaspamwherep ∤ m,thenasubgroup of G with order pa is called a Sylow p-subgroup of G, and is usually denoted by P . The number of Sylow p-subgroups of G will be denoted by n . p p Now for Sylow’s Theorem: Theorem 2.1.2. Let G be a group of order pam, where p ∤m. 3 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww 1. There is at least one Sylow p-subgroup of G 2. If P is any Sylow p-subgroup of G and Q is any p-subgroup of G, there exists g ∈ G such that Q ⊆ gPg−1, that is, every p-subgroup of G is contained in some conjugate of any Sylow p-subgroup, and in particular any two Sylow p-subgroups are conjugate 3. The number n of Sylow p-subgroups in G satisfies n ≡ 1 (mod p) and p p furthermore, n is equal to |G : N (P)|, the index of the normalizer p G of any Sylow p-subgroup in G First, we prove an auxiliary lemma about Sylow subgroups: Lemma 2.1.3. If P is a Sylow p-subgroup of G, and Q is any p-subgroup of G, then Q∩N (P) = Q∩P. G Proof. Write Q ∩ N (P) = H. Since P ⊆ N (P), certainly Q ∩ P ⊆ G G Q ∩ N (P), so we must only prove the reverse inclusion. Since clearly G H ⊆ Q, we must prove that H ⊆ P. Since H ⊆ N (P), we know that PH is a subgroup of G. We can also G see that PH is a p-group that contains P as a subgroup, so we can only have that PH = P, so indeed H ⊆ P. Now we are ready to prove Sylow’s theorem in full: Proof. We firstprove the existence of Sylow subgroupsof G by induction on its order. The base case is trivial, so assume that for a given group G, all groups of order less than G have Sylow subgroups. For a prime p dividing the order of G, we first suppose that p | ord(Z(G)). In this case, Cauchy’s theorem impliesthatZ(G)hasasubgroupN of orderp. Considerthegroup G/N, which has order pa−1m. The induction hypothesis implies that G/N has a Sylow p-subgroup P/N of order pa−1. But then P is a subgroup of G having order pa, so G also has a Sylow p-subgroup. Now, suppose that p ∤ ord(Z(G)). Let g ,g ,...g be representatives of 1 2 r thedistinct non-central conjugacy classes of G, and applythe class equation to get r ord(G) = ord(Z(G))+ |G : C (g )| X G i i=1 Since p |ord(G) but p ∤ord(Z(G)) by assumption, there is some i such that p ∤|G : C (g )|. For this i, ord(C (g )) = pak, with k < m since g ∈/ Z(G). G i G i i By the inductive hypothesis, this C (g ) has a subgroup of order pa, and G i therefore so does G. This proves 1. Now, for a given prime p dividing the order of a group G, we know that there exists some Sylow p-subgroup of G, which we denote by P. Let 4 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww S = {P ,P ,...,P } be the set of all conjugates of P in G, and let Q be 1 2 r any p-subgroup of G. The conjugation action of Q on S produces orbits s O ∪O ∪···∪O = S. Note that r = ord(O ). 1 2 s i P i=1 Without loss of generality, relabel the elements of S so that the first s conjugates are representatives of the orbits, so P ∈ O for 1 ≤ i ≤ s. For i i each i, we have that ord(O )ord(N (P )) = ord(Q). However, N (P ) = i Q i Q i N (P ) ∩ Q = P ∩ Q by Lemma 2.1.3, so ord(O ) = |Q : P ∩ Q| for G i i i i 1 ≤ i≤ s. Now, note that r is the number of conjugates of P in G, and is there- fore independent of our choice of group Q in the above conjugation action. Therefore, we may calculate r by taking the specific case Q = P . In this 1 case certainly ord(O ) = 1, while for i > 1, we must have ord(O ) = |P : 1 i 1 P ∩P | > 1. Furthermore, |P :P ∩P | must be divisible by p, since every 1 i 1 1 i s group in the expression is a p-group. Since r = ord(O )+ ord(O ), we 1 i P i=2 must have that r ≡1 (mod p). Let Q be any p-subgroup of G, and suppose that Q is contained in no conjugate of P. Then certainly Q ∩ P ( Q for all 1 ≤ i ≤ s, so i ord(O ) = |Q : Q ∩ P | is divisible by p for all 1 ≤ i ≤ s, contradicting i i r ≡ 1 (mod p). Thus, Q is contained in some conjugate of every Sylow p-subgroup of G, which also implies that every Sylow p-subgroup of G is conjugate. This proves 2. But now every Sylow p-subgroup of G is conjugate, and so there are precisely the r that were in our set S, so r = n , and n ≡ 1 (mod p). p p Furthermore, if all Sylow p-subgroups are conjugate, we must also have n = |G : N (P)| for any Sylow p-subgroup of G, and this proves 3. p G The following characteristics of Sylow subgroups are also important: Corollary 2.1.4. Let P be a Sylow p-subgroup of G. The following state- ments are equivalent: 1. P is the unique Sylow p-subgroup of G 2. P is normal in G 3. P is characteristic in G Proof. Suppose P is the unique Sylow p-subgroup of G. Then since the set ofconjugates ofP inGispreciselythesetofSylowp-subgroupsinG, wecan only have that gPg−1 = P for all g ∈ G, so P⊳G. Conversely, if P⊳G, then for any other Sylow p-subgroup Q of G, we have Q is conjugate to P, so for some g ∈ G we have Q = gPg−1 = P, so P is the unique Sylow p-subgroup 5 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww of G, and 1 and 2 are equivalent. However, characteristic implies normal, and uniqueness implies characteristic, so all of 1, 2, and 3 are equivalent. 2.2 Solvable Groups The results in this section are adapted from [5, p. 138]. Definition 2.2.1. The derived series of a group G is the series G = G ⊇ G ⊇ G ⊇ ... 0 1 2 where G is the commutator subgroup of G . A group G is said to be i+1 i solvable if its derived series terminates in the identity after a finite number of terms. For example, S has derived series S ,A ,(1), so S is solvable. The 3 3 3 3 smallest example of a nonsolvable group is A , and in general A is not 5 k solvable for any k ≥ 5. It is not hard to show that if G is solvable and H is normal in G that H and G/H are solvable. Of particular interest is that the converse is true: Lemma 2.2.2. Suppose that G is a group with a normal subgroup H, and suppose further that both H and G/H are solvable. Then G is solvable. Proof. Since G/H is solvable, its derived series is of the form G/H = G /H ⊇ G /H ⊇ ··· ⊇ {e} = H/H 0 1 On the other hand, this quotient series lifts to the first part of the derived series for G satisfying G= G ⊇ G ⊇ ··· ⊇ H 0 1 but since H is solvable, the remainder of the derived series for G terminates at the identity, so G is solvable. 2.3 Hall π-subgroups The results in this section are taken from [1, p.200] Definition 2.3.1. Let G be a finite group with order n, and let π be a set of primes dividing n. A Hall π-subgroup H of G is a subgroup such that every prime dividing |H| is in π, and no prime dividing |G :H| is in π. 6 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww As an example, the alternating group of five elements A has order 60 = 5 22×3×5, so A is a Hall {2,3}-subgroup of A . 4 5 While all possible Sylow p-subgroups always exist for any finite group G, Hall π-subgroups need not exist for all possible π. For example, while A has a Hall {2,3}-subgroup, it has neither a Hall {2,5}-subgroup nor a 5 Hall {3,5}-subgroup. However, if G is solvable, then it has all possible Hall subgroups: Theorem 2.3.2. If G isany finite solvable group and π is any setof primes, then G has a Hall-π subgroup. We first prove a very important lemma that will be used here and again in the section on supersolvable numbers: Lemma 2.3.3. Let G be a finite solvable group, and let N be a minimal normal subgroup of G (meaning that N ⊳G, and when there exists H ⊳G such that {e} ⊆ H ⊆ N, it must be that H = {e} or H = N). Then N must be elementary abelian (i.e. is an abelian group where every nonidentity element has the same prime order). Proof. Note that N is solvable since it is a subgroup of the solvable group G. If N = {e}, the claim is obvious. If N is nontrivial, we consider its commutator subgroup [N,N]. This commutator is characteristic in N, so since N ⊳G, we have [N,N]⊳G. Since N is minimal normal, it must be that [N,N] = {e} or [N,N] = N. However, [N,N] = N would imply that the derived series for N is a single repeating term that is not the identity, so then N would not be solvable, a contradiction. It must be then that [N,N] = {e}, which implies that N is abelian. Now let p be a prime that divides |N|. Then N has Sylow p-subgroups which are all conjugate to each other, but since N is abelian, conjugation is trivial, so there is a unique Sylow p-subgroup P. Then P is characteristic in N and N ⊳G, so P ⊳G. Clearly P is nontrivial, so P = N, and N is an abelian p-group. Finally, we let pN denote the set np|n ∈ N. It is easy to check that this is a subgroup of N, and that it is characteristic since it is invariant under isomorphisms of N. Again, we may conclude that pN = {e} or pN = N. However, Cauchy’s theorem tells us that N has at least oneelement of order p, so that pN is strictly contained in N. Therefore, pN = {e}, so every element of N must have order p, implying that N is elementary abelian. Now for the existence of Hall π-subgroups: Proof. Fix any set π of primes. We prove by induction that every finite solvable group G has Hall π-subgroups. In the base case, the trivial group is a Hall π-subgroup of itself for any π, and certainly our particular choice. 7 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww Now let G be a finite solvable group of order n, and suppose that every solvable group of order less than n has a Hall π-subgroup. Let N be some minimal normal subgroup of G. By the above lemma, N is elementary abelian, and in particular N is a p-group for some prime p. Since quotient groups of solvable groups are solvable, it must be that G/N is solvable, so by the inductive hypothesis, G/N has a Hall π-subgroup. If p ∈ π, then the Hall π-subgroup of G/N naturally corresponds to a Hall π-subgroup of G. Now we suppose that p is not in π. It is still true that G/N has a Hall π-subgroup K/N, and this corresponds to a subgroup K of G. Since subgroups of solvable groups are solvable, if K is a proper subgroup of G, then by the inductive hypothesis K has a Hall π-subgroup H. Now every prime dividing |H| is in π, but no prime dividing |G :K| or |K : H| is in π, so no prime dividing |G : K||K : H| = |G : H| is in π. It follows that H is a Hall π-subgroup of G. It remains to consider when p is not in π, and the Hall π-subgroup of G/N is itself. In this case, since N is a p-group and |G : N| is relatively prime to |N|, it must be that N is a Sylow p-subgroup of G. Let M/N be a minimal normal subgroup of G/N, which also corresponds to a normal subgroup M of G. It must be that M/N is of prime power order, but it cannot be of p-power order, since p ∤ |G : N|. Thus, M/N is an elementary abelian group of q-power order for some prime q 6= p. Note that p and q are then the only distinct prime factors dividing M. Also, q divides |G : N|, and since G/N is the Hall π-subgroup of itself, it must be that q ∈ π. Since q divides |M|, M has some Sylow q-subgroup Q. If Q⊳G, then G/Q has a Hall π-subgroup H/Q, but since Q is of q-power order and q ∈ π, the corresponding subgroup H of G is a Hall π-subgroup. We now prove Frattini’s Argument, which states that if M ⊳G and Q is a Sylow q-subgroup of M then G = MN (Q). To show this, consider the G subgroupgQg−1 formedbyconjugatingQwithanyg ∈ G. SinceQ ⊆ M and M⊳G, gQg−1 ⊆ M. SinceallSylowq-subgroupsofM areconjugate, itmust be that Q and gQg−1 are conjugate in M, so there must be some m ∈ M such that gQg−1 = mQm−1. Rearranging, we have that m−1gQg−1m = Q, so that m−1g ∈ N (Q). But then any g ∈ G may be written as m(m−1g), G a product of elements in M and N (Q), so G = MN (Q). G G Continuing the original proof, we may now assume that Q is not normal in G. By Frattini’s argument, G = MN (Q), and it follows that |G|×|M∩ G N (Q)| = |M|×|N (Q)|, or equivalently |G|/|N (Q)| = |M|/|M∩N (Q)|. G G G G Note that since Q ⊆ M ∩N (Q), we have that |M|/|M ∩N (Q)| divides G G |M|/|Q|. However, the only prime factors of |M| are p and q, so since Q is a Sylow q-subgroup of M, |M|/|Q| is a power of p, and then so is |M|/|M ∩N (Q)| = |G|/|N (Q)|. Since Q is not normal in G, N (Q) is G G G a proper subgroup of G that has a Hall π-subgroup H by the inductive hypothesis. Since |G|/|N (Q)| is a power of p ∈/ π, H is also a Hall π- G subgroup of G, completing the proof. 8 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww 3 Cyclic Numbers Lagrange’s theorem tells us that any prime number p must be a cyclic num- ber, but there are also many composite cyclic numbers. According to a well-known application of Cauchy’s theorem, if p and q are prime numbers with p > q and q ∤ p−1, then every group of order pq is cyclic ([6, p. 91]). This basic idea motivates the following theorem: Theorem 3.0.4. A positive integer n is a cyclic number ⇐⇒ (n,ϕ(n)) = 1. The following proof is from [3]: Proof. Suppose that there is some prime q such that q2 | n. Then since q | ϕ(qk) for k > 1, we have q | ϕ(n) and q | n, so certainly (n,ϕ(n)) 6= 1. Furthermore,thegroupC ×C isagroupofordernthatcannotbecyclic, q n/q as (q,n/q) 6= 1. For the remainder of the proof, we may assume that n is square-free, so n = p p ...p for distinct primes p ,p ,...,p . Supposethat (n,ϕ(n)) 6= 1. 1 2 r 1 2 r Then since n is square-free, we must have that p | p −1 for some primes i j p and p dividing n. Therefore, there exists a nontrivial homomorphism i j h : C → Aut(C ), and a corresponding nontrivial semidirect product pi pj C ⋊ C which produces a noncyclic group of order p p . It follows that pj h pi i j the direct product of this group with C is a group of order n which n/pipj is not cyclic, proving that (n,ϕ(n)) = 1 is necessary for n to be a cyclic number. To prove sufficiency, we proceed by contradiction. Let n be the smallest positive integer that is square-free and satisfies (n,ϕ(n)) = 1 but is not a cyclic number, and take G to be a group of order n that is not cyclic. Cauchy’stheoremtellsusthatthereexistelementsx inGfor1 ≤i ≤ rsuch i that ord(x ) = p for each i. If G were abelian, the product x = x x ...x i i 1 2 r would be an element of order n, so G would be cyclic, a contradiction. It is also clear that if d | n, then (n,ϕ(d)) = 1, and (d,ϕ(d)) = 1. In particular, this means that the minimality of n as a supposed counterexample implies if d | n and d < n, then d is a cyclic number, so it follows that all proper subgroups and quotient groups of G are cyclic. We now prove that Z(G) is trivial. Supposethat ord(Z(G)) > 1. In this case, the quotient group G/Z(G) must be cyclic as noted above. However, G/Z(G) cyclic implies that G is abelian, so Z(G) must be trivial. We claim that G has no normal subgroups other than G and {e}. Let H be a normal subgroup of G with H 6= G, and let d denote the order of H. Since H is proper in G, H must be cyclic. Let c : G → Aut(H) be 9 OOOnnn ttthhheee ccchhhaaarrraaacccttteeerrriiizzzaaatttiiiooonnn ooofff PPP nnnuuummmbbbeeerrrsss LLLooogggaaannn CCCrrreeewww the homomorphism induced by the conjugation action of G on H. The first isomorphism theorem for groups tells us that G/ker(c) is isomorphic to a subgroup of Aut(H), so its order divides ϕ(d). On the other hand, certainly |G/ker(c)| divides n, so since (n,ϕ(d)) = 1, we must have |G/ker(c)| = 1. Thus the kernel of the homomorphism c is the entire group G, so since c is conjugation, we have that every element of G commutes with every element of H. Therefore, H ⊆ Z(G), but since Z(G) is trivial, H must be trivial as well. Thus, if H ⊳G and H 6= G, it must be that H = {e}. We also claim that the intersection of any two distinct proper maximal subgroups of G is trivial. Let K and L denote two different proper maximal subgroups of G, and consider their intersection subgroup K ∩ L, and its centralizer C (K ∩L). By assumption, K and L are both cyclic, so both G abelian,implyingthatK andLbothcentralize K∩L. Therefore,C (K∩L) G contains both K and L, but since K and L are maximal and distinct, we can only have C (K ∩L) = G. Thus, K ∩L ⊂ Z(G), but since Z(G) is G trivial, so is K∩L. One immediate corollary of this is that the nonidentity elements of G are partitioned by the nonidentity elements of the distinct maximal subgroups of G, that is, every element in G\{e} is in exactly one of the M \{e}, where M ,M ,...,M are the maximal subgroups of G. i 1 2 k Now, we consider the action of conjugation by elements of G on the set of its maximal subgroups. Let k be the number of orbits created by the conjugation action, and let M denote a representative of each orbit, i for 1 ≤ i ≤ k. For notational simplicity, let ord(M ) = m . For each M i i i representing an orbit, we certainly have that the stabilizer of M contains i M . However, the stabilizer of M cannot be G, because we proved above i i that we cannot have M normal in G. Since M is maximal, we can only i i have that M is its own stabilizer, implying that the orbit it represents has i order n/m . i Therefore, we can use the partition established above to write that k n−1 = (n/m )(m −1) X i i i=1 because the number of nonidentity elements of G equals the number of nonidentity elements in the distinct maximal subgroups of G. Dividing by n, we obtain k 1−1/n = (1−1/m ) X i i=1 and expanding the summation, k 1−1/n = k− 1/m X i i=1 10