Home Character degree graphs of solvable groups with diameter three
Article Publicly Available

Character degree graphs of solvable groups with diameter three

  • Catherine B. Sass EMAIL logo
Published/Copyright: July 9, 2016

Abstract

Let G be a finite solvable group and cd(G) the set of character degrees of G. The character degree graph Δ(G) is the graph whose vertices are the primes dividing the degrees in cd(G) and there is an edge between two distinct primes p and q if their product pq divides some degree in cd(G). When Δ(G) has diameter three, we can partition the vertices ρ(G) into four non-empty disjoint subsets ρ1ρ2ρ3ρ4 where no prime in ρ1 is adjacent to any prime in ρ3ρ4; no prime in ρ4 is adjacent to any prime in ρ1ρ2; every prime in ρ2 is adjacent to some prime in ρ3; every prime in ρ3 is adjacent to some prime in ρ2; and |ρ1ρ2||ρ3ρ4|.

We will show the following: If G is a solvable group where Δ(G) has diameter three, then ρ3 has at least three vertices and G has a normal non-abelian Sylow p-subgroup where pρ3. If ρ1ρ2 has n vertices, then ρ3ρ4 must have at least 2n vertices. The group G has Fitting height 3.

1 Introduction

Let G be a finite solvable group, Irr(G) the set of irreducible characters and cd(G) the set of character degrees. We study solvable groups and their character degrees by studying Δ(G), the character degree graph. The set of vertices for Δ(G) is the set of all primes p where p divides a character degree. There is an edge between two distinct primes p and q if their product divides a character degree.

Many of the known results on character degree graphs can be found in the expository paper, [11]. One of the most important properties for Δ(G) when G is solvable is Pálfy’s Condition.

Theorem 1.1

Theorem 1.1 ([15])

A graph is said to satisfy Pálfy’s Condition if given any three vertices, two of them are adjacent. If G is a solvable group, then Δ(G) satisfies Pálfy’s Condition.

Two corollaries follow immediately. If Δ(G) is disconnected, there are at most two components, [12] and both components are complete. If Δ(G) is connected, the diameter is at most three [13].

There are two main questions that arise when studying character degree graphs. What can we say about the group or family of groups that have a particular character degree graph? Which graphs can occur as a character degree graph? The following three theorems answer the second question for the graphs with five vertices or fewer, and give a partial answer for the graphs with six vertices.

Theorem 1.2

Theorem 1.2 ([17])

The graph in Figure 1 is not the character degree graph of any solvable group.

Figure 1 Graph with four vertices and diameter three.
Figure 1

Graph with four vertices and diameter three.

Theorem 1.3

Theorem 1.3 ([9])

The graphs in Figure 2 are not the character degree graphs of any solvable group.

Figure 2 Graphs with five vertices and diameter three.
Figure 2

Graphs with five vertices and diameter three.

A consequence of Theorems 1.2 and 1.3 is that when Δ(G) has diameter three, it must have at least six vertices. In fact, there is a family of solvable groups that have a character degree graph with diameter three and six vertices. Lewis gave an example in [8] and Dugan generalized in [2] that result to show that a family of groups have a character degree graph with diameter three.

Theorem 1.4

Theorem 1.4 ([8])

There exists a solvable group that has a character degree graph as shown in Figure 3.

Figure 3 Graph with six vertices and diameter three.
Figure 3

Graph with six vertices and diameter three.

In [11], the question was posed, do the subgraphs shown in Figure 4, of the graph in Figure 3, occur as Δ(G) where G is a solvable group? In this paper we show that in fact, the graphs that are isomorphic to those in Figure 4 do not occur as Δ(G) for any solvable group G.

Figure 4 Graphs with six vertices and diameter three.
Figure 4

Graphs with six vertices and diameter three.

Let G be a solvable group and Δ(G) the character degree graph that has diameter three. To easily describe the graph Δ(G), we define a partition on the set of vertices ρ(G) as Lewis did in [10]. Because Δ(G) has diameter three, we can find two vertices distance three from each other. Label them p1 and p4. Because p1 and p4 are necessarily not adjacent, if we consider any other prime qρ(G), the prime q must be adjacent to either p1 or p4. We define the following four sets:

Definition 1.5

  1. Define ρ4(G) to be the set of all vertices that are distance three from the vertex p1. As p4 is distance three from p1, the vertex p4ρ4.

  2. Define ρ3(G) to be the set of all vertices that are distance two from the vertex p1.

  3. Define ρ2(G) to be the set of all vertices that are adjacent to p1 and adjacent to some prime in ρ3(G).

  4. Define ρ1(G) to be the vertices consisting of p1 and those that are adjacent to p1 and not adjacent to anything in ρ3(G).

  5. We relabel if necessary so that |ρ1ρ2||ρ3ρ4|

With the above partition in mind, we state our main theorems:

Theorem 1

The graphs in Figure 4 do not occur as Δ(G) for any solvable group G. In particular, if G is a solvable group where Δ(G) has diameter three and |ρ(G)|=6, then Δ(G) is isomorphic to the graph in Figure 3.

We are able to give a lower bound on the number of vertices in the subset ρ3.

Theorem 2

Let G be a solvable group where Δ(G) has diameter three. Then |ρ3|3.

Theorem 3

Let G be a solvable group where Δ(G) has diameter three. Then G has a normal non-abelian Sylow p-subgroup for exactly one prime p and pρ3.

In order to prove Theorem 3, we use the previous results but we also need to show more generally that G would not have a normal non-abelian Sylow p-subgroup where p is in the subset ρ4. Finally, we confirm [11, Conjecture 4.8] and give the Fitting height of all solvable groups that have a character degree graph with diameter three.

Theorem 4

Let G be a solvable group with Δ(G) having diameter three. If n=|ρ1ρ2|, then |ρ3ρ4|2n.

Theorem 5

Let G be a solvable group with Δ(G) having diameter three. Then G has Fitting height 3.

2 Background

Let M be a normal subgroup of G. The graphs Δ(M) and Δ(G/M) must be subgraphs of Δ(G). If ρ(M) or ρ(G/M) contains primes from both ρ1(G) and ρ4(G) then the graph Δ(M) or Δ(G/M) must either have diameter three or be disconnected. Because we often know that |G| is minimal with Δ(G) having diameter three, we rely heavily on the classification of disconnected character degree graphs in [7]. We state the theorem here for convenience.

Theorem 2.1

Theorem 2.1 ([7])

Let G be a solvable group where Δ(G) has two connected components. Then G is one of the following examples:

  1. G has a normal non-abelian Sylow p-subgroup P and an abelian p-complement K for some prime p. The subgroup PP(K) and every non-linear irreducible character of P is fully ramified with respect to P/P(K).

  2. G is the semi-direct product of a subgroup H acting on a subgroup P, where P is elementary abelian of order 9 and cd(H)={1,2,3}. Let Z=H(P). We have Z(H), and H/ZSL2(3), where the action of H on P is the natural action of SL2(3) on P.

  3. G is the semi-direct product of a subgroup H acting on a subgroup P, where P is elementary abelian of order 9 and cd(H)={1,2,3,4}. Let Z=H(P). We have Z(H), and H/ZGL2(3), where the action of H on P is the natural action of GL2(3) on P.

  4. G is the semi-direct product of a subgroup H acting on an elementary abelian p-group V for some prime p. Let Z=H(V) and K the Fitting subgroup of H. Write m=|H:K|>1, and |V|=qm, where q is a p-power. We have Z(H), K/Z is abelian, K acts irreducibly on V, m and |K:Z| are relatively prime, and (qm-1)/(q-1) divides |K:Z|.

  5. G has a normal non-abelian 2 -subgroup Q, and an abelian 2 -complement K with the property that |G:KQ|=2 and the quotient G/Q is not abelian. Let Z=K(Q), and C=Q(K). The subgroup QC, and Z is central in G. Every non-linear irreducible character of Q is fully ramified with respect to Q/C. Furthermore, Q/C is an elementary abelian 2 -group of order 22a for some positive integer a, Q/C is irreducible under the action of K, and K/Z is abelian of order 2a+1.

  6. G is the semi-direct product of an abelian group D acting coprimely on a group T so that [T,D] is a Frobenius group. The Frobenius kernel is A=T=[T,D], A is a non-abelian p-group for some prime p, and a Frobenius complement is B with [B,D]B. Every character in Irr(TA) is invariant under the action of D and A/A is irreducible under the action of B. If m=|D:D(A)|, then |A:A|=qm where q is a p-power, and (qm-1)/(q-1) divides |B|.

Two of these families never show up in our work. They are Examples (2.2) and (2.3). They both have two connected components each with one vertex, two and three. They have no normal non-abelian Sylow p-subgroups. When we are considering a graph that has two connected components each with a singleton vertex, one of those vertices, p, corresponds to a normal non-abelian Sylow p-subgroup.

Two of the remaining families have no normal non-abelian Sylow p-subgroups. They are Examples (2.4) and (2.5). Example (2.5) has a character degree graph where the smaller component is the singleton 2. Examples (2.1) and (2.6) both have a normal non-abelian Sylow p-subgroup.

We use the Zsigmondy Prime Theorem to count primes or vertices in a disconnected graph. Let q and n be positive integers. A prime p is called a Zsigmondy prime divisor for qn-1 if p divides qn-1 and p does not divide qj-1 for 1j<n. The Zsigmondy Prime Theorem says there exists a Zsigmondy prime for qn-1 unless either n=2 and q=2k-1 for some integer k, or n=6 and q=2. Because of these exceptional cases, we use the following lemma from [7] which takes care of the exceptional cases.

Lemma 2.2

Lemma 2.2 ([7, Lemma 5.1])

Let m be a positive integer, and let q be a prime power such that (qm-1)/(q-1) is relatively prime to m. If r is the number of distinct prime divisors of m, then the quotient (qm-1)/(q-1) has at least 2r-1 prime divisors.

Another result by Pálfy is used to count vertices in the larger component when the graph is disconnected. We call this result Pálfy’s inequality.

Theorem 2.3

Theorem 2.3 ([16, Theorem 3])

Let G be a solvable group with a disconnected character degree graph Δ(G). If the smaller component has n vertices, then the larger component has at least 2n-1 vertices.

In Examples (2.4) and (2.6) from Theorem 2.1, let F=𝔽(G), E/F=𝔽(G/F) and m=|G:E|. The prime divisors of m are precisely the primes in the smaller component of the character degree graph. Let r be the number of distinct primes that divide m. We know from Pálfy’s inequality that we must have at least 2r-1 primes in the larger component. By [7, Theorem 5.4], we know that if the larger component has exactly 2r-1 primes, then F is abelian and in particular, we are in Example (2.4). When there are more than 2r-1 primes in the larger component, we may have to distinguish between Examples (2.4) and (2.6). Of course, if the group has a normal non-abelian Sylow p-subgroup P, then we are in Example (2.6). In this case, G/P will satisfy Example (2.4) by [7, Lemma 3], and so the larger component in Δ(G) must have at least 2r primes including p.

We know from Ito’s Theorem [3, Corollary 12.34] that G has a normal abelian Sylow p-subgroup if and only if pρ(G). If the prime or vertex p is in ρ(G), we know that if the Sylow subgroup is normal, it must be non-abelian. It is redundant to keep saying that the Sylow subgroup is non-abelian, but we do so anyway to avoid confusion. With that in mind, this lemma from [9] is used frequently.

Lemma 2.4

Lemma 2.4 ([9, Lemma 3])

Suppose that G has a non-abelian normal Sylow p-subgroup P for a prime p. Then ρ(G/P)=ρ(G)\{p}.

One of our goals is to show that if G is a solvable group with a character degree graph having diameter three, then G has exactly one normal non-abelian Sylow p-subgroup and pρ3. This result from [18] tells us that we can have at most one normal non-abelian Sylow p-subgroup when Δ(G) has diameter three.

Theorem 2.5

Theorem 2.5 ([18])

If G is a solvable group and Δ(G) has diameter three, then G has at most one normal non-abelian Sylow p-subgroup for some prime pρ(G).

Finally, this theorem from [9] gives us our strategy for proving that the graphs we are considering are not character degree graphs for any solvable group G.

Theorem 2.6

Theorem 2.6 ([9, Theorem 2])

Suppose G is a solvable group with Φ(G)=1. Assume that for all nontrivial normal subgroups M, Δ(G/M) has two connected components or the diameter of Δ(G/M) is at most 2. Then either Δ(G) has two connected components or the diameter of Δ(G) is at most 2.

3 Observations of the partition in Definition 1.5

Let G be a solvable group and Δ(G) the character degree graph with diameter three such that the vertices ρ(G)=ρ1ρ2ρ3ρ4.

  1. Every prime in ρ(G) is in exactly one of the four sets.

  2. The sets ρ2 and ρ3 are non-empty.

Proof.

Because there is a shortest path between p1 and p4, there exists primes p2 and p3 such that the path p1,p2,p3,p4 is a shortest path between p1 and p4. The prime p3 is necessarily in ρ3 as p3 is distance 2 from p1 otherwise it contradicts that p1,p2,p3,p4 is a shortest path. The prime p2 is necessarily in ρ2 as p2 is adjacent to something in ρ3 and also adjacent to p1. ∎

  1. Because no prime in ρ3ρ4 is adjacent to the prime p1, the subset ρ3ρ4 determines a complete subgraph of Δ(G).

  2. Because no prime in ρ1ρ2 is adjacent to the prime p4, the subset ρ1ρ2 determines a complete subgraph of Δ(G).

  3. Every prime in ρ2 is adjacent to some prime in ρ3 and every prime in ρ3 is adjacent to some prime in ρ2.

  4. Every prime in ρ3 is distance 2 from any prime in ρ1.

  5. The subset ρ2 is the set of vertices that are distance 2 from any prime in ρ4.

  6. The subset ρ1 is the set of all vertices that are distance 3 from any prime in ρ4.

Proof.

Let r be a prime in ρ1 and s be a prime in ρ4. Because ρ1ρ2 determines a complete subgraph of Δ(G) and ρ2 is non-empty, there is a prime p2ρ2 that is adjacent to the prime r. Because p2 is adjacent to some prime in ρ3 and ρ3 is non-empty, there exists a prime p3ρ3 where p2 and p3 are adjacent. If p2 were adjacent to the prime s, then in particular, s would be distance two from p1 and s would be in ρ3. This contradicts the fact that sρ4. As ρ3ρ4 determines a complete subgraph of Δ(G), the prime p3 is adjacent to s and we have found a shortest path r,p2,p3,s. ∎

  1. If r and s are primes in ρ(G) that are distance three from each other, than one must be in ρ1 and the other must be in ρ4.

Proof.

Suppose that r is a prime that is not in ρ1 or ρ4. Without loss, we can assume that r is in ρ2. Because ρ1ρ2 determines a complete subgraph, we see that s cannot be in ρ1. Thus s must be in either ρ3ρ4. Because r is adjacent to some prime in ρ3, there exists a p3ρ3 and r and p3 are adjacent. Because ρ3ρ4 determines a complete subgraph, we see that p3 is adjacent to s and the distance between r and s must be less than or equal to two. ∎

  1. If the partition is defined using r and s instead of p1 and p4, where the distance between r and s is three, the partition will be the same.

  2. Further, the partition is unique up to symmetry in the case where |ρ1ρ2|=|ρ3ρ4|. We will show that it is the case that |ρ1ρ2|<|ρ3ρ4|, and the partition for character degree graphs will be unique.

The following theorem shows that the only possible graphs with six vertices that satisfy Pálfy’s condition that can be character degree graphs are those in Figure 3 and Figure 4 and in particular, the partition defined in Definition 1.5 is unique when |ρ(G)|=6.

Theorem 3.1

Theorem 3.1 ([9, Theorem 4])

If |ρ(G)|=6, then

|ρ1ρ2|=2𝑎𝑛𝑑|ρ3ρ4|=4.

4 Preliminary results

In this section, we prove many of the lemmas necessary to show that the graphs in Figure 4 do not occur as character degree graphs for any solvable group G. The proof that the graphs in Figure 4 do not occur follows in the next section.

The following lemma shows that if Δ(Oρ2(G)) is a disconnected graph, then Oρ2(G) cannot be Example (2.4) or Example (2.5) from Theorem 2.1.

Lemma 4.1

Let G be a solvable group with Δ(G) having diameter three. Assume that Oρ2(G)<G and that the graph Δ(Oρ2(G)) does not have diameter three. Then, Oρ2(G) must have a normal non-abelian Sylow p-subgroup and the graph Δ(Oρ2(G)) is disconnected. In particular, the group Oρ2(G) is not Example (2.4) or Example (2.5) from Theorem 2.1.

Proof.

Suppose that G is a solvable group and Δ(G) has diameter three. Let K=Oρ2(G) and suppose K<G and so Δ(K) is not diameter three. We assume that K has no normal non-abelian Sylow p-subgroups and derive a contradiction.

Claim

The subgroup K is Example (2.4) or Example (2.5) from Theorem 2.1.

Proof.

Because G/K is a ρ2-group, all primes in ρ(G)\ρ2 are contained in ρ(K). Because the primes in ρ1 and ρ4 are contained in ρ(K) and Δ(K) is a subgraph of Δ(G), we have Δ(K) is either disconnected or has diameter three. By the hypothesis, Δ(K) does not have diameter three, and so it must be disconnected. If pρ(K) for pρ2, then since |ρ1|1, the smaller component has at least two primes. By [7, Theorem 5.3], Op(K)<K. This contradicts the definition of K, and so ρ2 intersects ρ(K) trivially. Thus, the two components of Δ(K) are ρ1 and ρ3ρ4.

If K has a normal Sylow p-subgroup for some prime pρ(K), so does G as ρ2 intersects ρ(K) trivially. As G has no normal non-abelian Sylow p-subgroups, K also has no normal Sylow p-subgroups. Because Δ(K) is disconnected, the subgroup K is either Example (2.4) or Example (2.5) from Theorem 2.1. ∎

Thus K is Example (2.4) or Example (2.5) from Theorem 2.1 with the larger component being ρ3ρ4. By [7, Theorem 5.2], cd(K) contains only one degree that is divisible by primes in ρ3ρ4. Pick primes p2ρ2, p3ρ3, and p4ρ4 such that p2 and p3 are adjacent in Δ(G). Then there exists a character χIrr(G) such that p2p3 divides χ(1). Let θ be an irreducible constituent of χK in Irr(K). By [3, Corollary 11.29], we know χ(1)/θ(1) divides |G:K|, and hence, p3 divides θ(1). Because K has only one degree divisible by primes in ρ3ρ4 and that degree is divisible by p3 and p4, we conclude that p4 also divides θ(1). Thus, p2 and p4 are adjacent in Δ(G), which is a contradiction. ∎

The next two lemmas follow from the same arguments given in Claim 3 and Claim 5 in the proof of Theorem 1.3 in [9].

Lemma 4.2

Let G be a solvable group with a normal Sylow p-subgroup P for some prime pρ(G). Suppose ρ(G/M)=ρ(G) implies that M=1 whenever M is a normal subgroup of G. Then P is a minimal normal subgroup in G and in particular, P is a central subgroup in P.

Proof.

Let G be a solvable group with a normal Sylow p-subgroup P for some prime pρ(G). Assume ρ(G/M)=ρ(G) implies M=1. By Lemma 2.4, we have ρ(G/P)=ρ(G)\{p}. Let X be a normal subgroup of G that is contained in P so that P/X is a chief factor for G. Because ρ(G/X) contains p and ρ(G/P), we see that ρ(G/X)=ρ(G). Thus X=1 by our hypothesis and P is a minimal normal subgroup of G. Because (P) is a characteristic subgroup of P, it is normal in G. It follows that P(P) is a normal subgroup of G. Because P is a nilpotent subgroup of G, P(P) is not a trivial subgroup, and so P(P). ∎

Lemma 4.3

Let G be a solvable group and suppose G has a normal Sylow p-subgroup for some prime pρ(G). Suppose there is a normal subgroup N in G such that p does not divide |N|. Then ρ(G/N) contains every prime in ρ(G) that is not adjacent to p in Δ(G).

Proof.

Let G be a solvable group with a normal Sylow p-subgroup P for some prime pρ(G). Let N be a normal subgroup of G such that p does not divide |N|. The subgroup PN/N is a normal non-abelian Sylow p-subgroup of G/N and so pρ(G/N). Let q be a prime in ρ(G)\ρ(G/N), and let Q be a Sylow q-subgroup of G. Since q divides no degree in cd(G/N), we use Itô’s Theorem [3, Corollary 12.34] to see that QN/N is abelian and QN is normal in G. Thus, the direct product P×QN is normal in G. Let χIrr(G) with q dividing χ(1) and let θIrr(QN) be an irreducible constituent of χQN. We know that χ(1)/θ(1) divides |G:QN| by [3, Corollary 11.29]. Since q does not divide |G:QN|, we see q must divide θ(1). Thus, qρ(QN), and so p and q will be adjacent in Δ(P×QN). Because Δ(P×QN) is a subgraph of Δ(G), the primes p and q are adjacent in Δ(G). In particular, ρ(G/N) contains every prime in ρ(G) that is not adjacent to p in Δ(G). ∎

Because we know that a solvable group G has at most one normal non-abelian Sylow p-subgroup, if Δ(G) has diameter three, then all factor groups G/M have restrictions on the possible normal Sylow p-subgroups. This is the case even if Δ(G/M) does not have diameter three.

Lemma 4.4

Let G be a solvable group where Δ(G) has diameter three. Let M be a minimal normal subgroup of G such that Δ(G/M) has diameter three. If G has a normal Sylow p-subgroup for pρ(G), then G/M does not have a normal Sylow q-subgroup for qρ(G)\{p}.

Proof.

Let G be a solvable group with Δ(G) having diameter three. Suppose G has a normal Sylow p-subgroup P for pρ(G), and let M be a minimal normal subgroup of G and suppose Δ(G/M) has diameter three. Suppose that the subgroup G/M has a normal non-abelian Sylow q-subgroup Q/M for a prime qρ(G)\{p}. Because the subgroup PM/M is a normal Sylow p-subgroup of G/M, the Sylow subgroup PM/M must be abelian or we have a contradiction as G/M can have at most one normal non-abelian Sylow p-subgroup. Thus, PM. Let R be a Sylow q-subgroup of G. We see that RM is normal in G and so RP is a normal subgroup in G. Further, [R,P]M and so R centralizes P/P by [5, Corollary 3.28]. Hence, P=P(R)P=P(R)Φ(P). As PΦ(P), by the Frattini Argument [4], P=P(R) and so RP is a direct product. Thus R is normal in RP, so it is characteristic, and R is normal in G. This contradicts the fact that G can have at most one normal non-abelian Sylow p-subgroup. ∎

We will eventually show that if Δ(G) has diameter three, then G has a normal Sylow p-subgroup for some prime pρ3. However, if ρ3 is not large enough, the solvable group G cannot have a normal Sylow p3-subgroup and a diameter three character degree graph. This is necessary to show that the graphs in Figure 4 do not exist as character degree graphs.

Lemma 4.5

Let G be a solvable group with Δ(G) having diameter three. Assume for all proper nontrivial normal subgroups M that Δ(G/M) and Δ(M) do not have diameter three, and assume |ρ3|<2n-1 where n=|ρ1ρ2|. Then G does not have a normal Sylow p3-subgroup for any prime p3ρ3.

Proof.

Suppose G has a normal non-abelian Sylow p3-subgroup P for a prime p3ρ3. Then, since ρ(G/P)=ρ(G){p3} by Lemma 2.4 and the group G/P is a nontrivial proper factor group of G, the graph Δ(G/P) cannot have diameter three by the hypothesis. Because ρ(G/P) contains primes from ρ1 and ρ4, we have that the graph Δ(G/P) must be disconnected with components ρ1ρ2 and (ρ3ρ4){p3}. Further, G/P cannot have any normal non-abelian Sylow subgroups or G would have more than one normal non-abelian Sylow subgroup, and that violates Theorem 2.5. So G/P is Example (2.4) from Theorem 2.1 as ρ1ρ2ρ(G/P) and |ρ1ρ2|2.

Theorem 5.3 in [7] gives us that Op2(G/P)<G/P, and so Op2(G)<G for every prime p2ρ2. Fix the prime p2ρ2 and let K=Op2(G). Because ρ(K) contains ρ1 and ρ4, and K is a proper subgroup of G, the graph Δ(K) is a subgraph of Δ(G). By the hypothesis, the graph Δ(K) must be disconnected. As K contains the subgroup P, the subgroup K has a normal non-abelian Sylow p3-subgroup and p3 is in the larger component. So K must satisfy the hypotheses of Example (2.6) from Theorem 2.1.

By [7, Lemma 3.6], K/P satisfies the hypotheses of Example (2.4). Let

S/P=𝔽(K/P),R/S=𝔽(K/S),
F/P=𝔽(G/P),E/F=𝔽(G/F).

Because S is a characteristic subgroup of K and K is normal in G, S is normal in G, and SFK. Because FK is a normal subgroup of K, we have FK=S. Further, R is characteristic in K and so R is normal in G. By the Diamond Lemma in [5], we obtain R/SRF/F is a normal subgroup of G/F. Because RF/F is normal in G/F, we have RF/FE/F and RFE. Since E/F is a Hall (ρ3ρ4){p}-subgroup, and R/S is a Hall (ρ3ρ4){p3}-subgroup by [7, Lemma 3.4], RF=E.

Because G/P satisfies the hypotheses of Example (2.4), G/P is the semi-direct product of a subgroup HP/P acting on an elementary abelian p3-group P/P, the positive integer m is defined to be the index |H:𝔽(H)|, the order of P/P is qm where q is a p3-power, and the quotient (qm-1)/(q-1) divides |E:F|. Since there are |ρ1ρ2|=n primes that divide m, there must be 2n-1 distinct primes that divide the quotient (qm-1)/(q-1) by Lemma 2.2. Because p3 does not divide |E:F|, and there are less than 2n-1 primes in ρ3, there must be at least one prime p4ρ4 that divides (qm-1)/(q-1).

As the graph Δ(G) is connected, there exist primes p2ρ2 and p3ρ3 where p2 and p3 are adjacent. Then there is a character χIrr(G) such that p2p3 divides χ(1). Let θ be an irreducible constituent of χK. By [3, Corollary 11.29], p3 divides θ(1) and so P is not contained in kerθ. The size of the smaller component of Δ(K) is either |ρ1ρ2|=n or |(ρ1ρ2)\{p2}|=n-1=a.

Because K is Example (2.6), K contains a Frobenius group where P is the Frobenius kernel and we let B be a Frobenius complement. The positive integer m1 is defined to be the index |K:R|. There is a p3-power q1 such that |P:P|=q1m1 and (q1m1-1)/(q1-1) divides |B|. By [7, Lemma 3.6], the character degrees cd(KP) are all divisible by p|B|. Because θIrr(KP), the quotient (q1m1-1)/(q1-1) divides θ(1).

If the smaller component of Δ(K) is ρ1ρ2, then

m=m1and(q1m1-1)/(q1-1)=(qm-1)/(q-1).

Since the prime p4 divides (qm-1)/(q-1), the prime p4 divides θ(1). If the smaller component of Δ(K) is ρ1ρ2\{p2}, then let m2=|G:K|. Then

m1=m/m2andq1m1=qm=|P:P|,

so q1=qm2. Let r be a Zsigmondy prime divisor of qm. Then r divides qm-1 and r does not divide qs-1 for any s<m. Because q1m1-1=qm-1, and m2<m, the prime r divides the quotient (q1m1-1)/(q1-1). Thus, as p4 is a Zsigmondy prime divisor of qm-1, the prime p4 divides (q1m1-1)/(q1-1) and so divides θ(1).

As p4 divides θ(1), the prime p4 also divides χ(1). Then the primes p2 and p4 are adjacent, but this is impossible as p2ρ2 and p4ρ4. Thus, G has no normal Sylow p3-subgroup. ∎

The following lemma shows that under suitable hypotheses, if we have a nontrivial factor group G/M, then |ρ(G/M)|<|ρ(G)|.

Lemma 4.6

Let G be a solvable group such that Δ(G) has diameter three. Assume for all proper nontrivial normal subgroups N that Δ(G/N) and Δ(N) do not have diameter three. Further, assume G has no normal Sylow p4-subgroup for p4ρ4, Oρ4(G)=G, |ρ1ρ2|=n, and |ρ3|<2n-1. Then ρ(G/M)=ρ(G) implies M=1, whenever M is a normal subgroup of G.

Proof.

Suppose G is a solvable group where Δ(G) has diameter three, G has no normal Sylow p4-subgroups for any prime p4ρ4, the subgroup Oρ4(G)=G, |ρ1ρ2|=n, and |ρ3ρ4|<2n-1. Assume that if N is a proper nontrivial normal subgroup of G that the graphs Δ(N) and Δ(G/N) do not have diameter three. Let M be a nontrivial normal subgroup of G, and suppose that ρ(G/M)=ρ(G). Then Δ(G/M) does not have diameter three, and since Δ(G/M) is a subgraph of Δ(G), the graph Δ(G/M) must be disconnected with components ρ1ρ2 and ρ3ρ4. The group G/M must be one of the examples from Theorem 2.1. Because both components have more than one vertex, we must be in Example (2.4) or Example (2.6). If G/M is Example (2.4), then it has no normal Sylow p-subgroups for any prime pρ(G/M) and in particular, G has no normal non-abelian Sylow p-subgroups either as ρ(G/M)=ρ(G). By [7, Theorem 5.3], we have Op2(G/M)<G/M. Let H/M=Op2(G/M). Then since |G/M:H/M| is a nontrivial power of p2, the index |G:H| is a nontrivial power of the prime p2. So Op2(G)<G. This contradicts Lemma 4.1, hence G/M is Example (2.6). By [7, Lemma 3.6], the group G/M has a normal Sylow p-subgroup for some prime pρ3ρ4. Let Q/M be that Sylow p-subgroup of G/M and P a Sylow p-subgroup of G that is contained in Q.

Consider K=Op2(G) for a fixed prime p2ρ2. Because Op2(G/M)<G/M by [7, Theorem 5.3], we have K is a proper subgroup of G, and so Δ(K) must be disconnected. Also, notice P is contained in K.

Claim

The group K is Example (2.6) from Theorem 2.1.

Proof.

Because the only prime that could be missing from ρ(K) is the prime p2, at least one of the components is larger than one as ρ(G)\{p2}ρ(K). If K is Example (2.1) or Example (2.5), then |ρ1|=|ρ2|=1, the prime p2ρ(K), and K has an abelian p1-complement. However, this implies Op4(K) is a proper subgroup of K. Then Op4(G) is a proper subgroup in G and this is a contradiction.

Suppose K is Example (2.4). There exists a prime p3ρ3 that is adjacent to p2 and a character χIrr(G) such that p2p3 divides χ(1). Let θ be an irreducible constituent of χK. Then by [3, Corollary 11.29], the quotient χ(1)/θ(1) divides the index |G:K|. Because the index |G:K| is a p2-power, p3 divides θ(1). Because K is Example (2.4), there is only one character degree that is divisible by the primes in the larger component, ρ3ρ4, and so θ(1) must be that character degree. Hence, every prime p4ρ4 divides θ(1). Thus, p4 divides χ(1) and p2 and p4 are adjacent, which is a contradiction. Thus, K must be Example (2.6) from Theorem 2.1. ∎

Since K is Example (2.6), K has a normal non-abelian Sylow r-subgroup R for some prime rρ3ρ4. Because R is a characteristic subgroup of K, R is normal in G. Further, as |G:K| is a power of p2, R is a Sylow subgroup of G. Because G/M has a normal non-abelian Sylow p-subgroup, and RM/M is a normal non-abelian Sylow r-subgroup of G/M, p=r and R=P. As G has no normal non-abelian Sylow p-subgroups for pρ4, the prime p must be in ρ3. But this contradicts Lemma 4.5. Thus ρ(G/M) does not equal ρ(G). ∎

5 Graphs with six vertices

In this section we show that if G is a solvable group then Δ(G) is not one of the two graphs from Figure 4. We start by showing that under suitable conditions, if Δ(G) has diameter three, then G does not have a normal non-abelian Sylow p-subgroup for any prime pρ1ρ2.

Lemma 5.1

Let G be a solvable group with Δ(G) having diameter three. Assume for all proper nontrivial normal subgroups M that Δ(G/M) and Δ(M) do not have diameter three. Further, assume ρ(G/M)=ρ(G) implies M=1. Then G does not have a normal Sylow p2-subgroup for any prime p2ρ2.

Proof.

Suppose G has a normal Sylow p2-subgroup P for a fixed prime p2ρ2 and let H be a p2-complement of G. By [9, Lemma 3], ρ(G/P)=ρ(G)\{p2}. Because ρ(G/P) contains the primes in ρ1 and ρ4, and Δ(G/P) is a subgraph of Δ(G), the graph Δ(G/P) is disconnected. Because P is not central in G and H acts on P nontrivially, we see that P/P is not central in G/P. From [7, Theorem 5.5], the Fitting subgroup of G/P has at most one non-central Sylow subgroup, which is P/P. Let F=𝔽(G). Since all of the other Sylow subgroups of 𝔽(G/P)=F/P are central, F/P is abelian, and thus G/P is as described in Example (2.4) from Theorem 2.1. This is because any solvable group having an abelian Fitting subgroup whose graph has two connected components, where at least one connected component has size larger than one, must satisfy the hypotheses of Example (2.4). Further, G/P has Fitting height 3.

Let E/F=𝔽(G/F) and Z=H(P/P). Then the quotient G/E is a cyclic (ρ1ρ2)\{p2}-group, E/F is a cyclic ρ3ρ4-group and EH is abelian. Because H/(EH) is isomorphic to G/E, the subgroup EH contains every Sylow p4 subgroups of G as G/E is a ρ1ρ2-subgroup. So, the subgroup H has a normal Sylow p4-subgroup Q for some p4ρ4. By [7, Lemma 3.4] we have that EH acts irreducibly on [E,F]/P. Also, ρ(PQ)={p2,p4} and Δ(PQ) has two connected components. Because the Fitting height of PQ is 2, by [7, Lemma 4.1] we know that PQ is Example (2.1) of Theorem 2.1. Let C=P(Q). Because PQ is Example (2.1) from Theorem 2.1, PC, every non-linear irreducible character is fully ramified with respect to P/C, and Q acts nontrivially on P fixing every non-linear irreducible character of P. By [14, Theorem 19.3], we have P=[P,Q], and so P<[P,Q]. Because PF and QE, we have [P,Q][E,F]. Since [E,F]/P is irreducible under the action of EH, we have [E,F]=[P,Q].

Fix a prime p1ρ1. As p1 and p2 are adjacent in Δ(G), there exists a character χIrr(G) such that p1p2 divides χ(1). Let θIrr(P) be an irreducible constituent of χP. Notice that θ is invariant in F as F is nilpotent, and in E because no prime in ρ3ρ4 divides χ(1). By Glauberman’s Lemma [3, Lemma 13.8], there exists an EH-invariant irreducible constituent of θP. Thus, P(EH)>1. By Lemma 4.2, P is central in P and so P(EH) is normal in P. Also, H normalizes P and EH. So, H and G=PH normalize P(EH). Furthermore, P is minimal normal in G by Lemma 4.2, thus, PP(EH). By Fitting’s Lemma,

P/P=[P,Q]/P×C/P=[E,F]/P×P(EH)/P.

Because P(EH)C, we have P(EH)=C and so E satisfies the hypotheses of Example (2.1) and has components {p2} and ρ3ρ4 by [7, Lemma 3.1]. There exists a character ψIrr(G) such that p2p3 divides ψ(1) for some p3ρ3. Let γIrr(E) such that γ is an irreducible constituent of ψE. By [3, Corollary 11.29], the quotient ψ(1)/γ(1) divides |G:E|, and since neither p2 nor p3 divides |G:E|, we have that p2 and p3 divide γ(1). This is a contradiction to the fact that Δ(E) is disconnected. Thus, G has no normal Sylow p2-subgroup. ∎

The hypotheses for the following lemma have the additional condition that Op(G)=G for all primes pρ3ρ4.

Lemma 5.2

Let G be a solvable group with Δ(G) having diameter three. Assume for all proper nontrivial normal subgroups M that Δ(G/M) and Δ(M) do not have diameter three. Further, suppose that ρ(G/M)=ρ(G) implies M=1 and Oρ3(G)=G or Oρ4(G)=G. Then G does not have a normal Sylow p1-subgroup for any prime p1ρ1.

Proof.

Suppose G has a normal Sylow p1-subgroup P for some p1ρ1, and let N=Op1(G). By Lemma 4.3, we have {p1}ρ3ρ4ρ(G/N). By the hypotheses, if N>1, then |ρ(G/N)|<|ρ(G)|, and so, by Lemma 4.3, there exists a prime p2(ρ1ρ2)\{p1} that is not in ρ(G/N). Since ρ(G/N) contains {p1} and ρ4, it follows that Δ(G/N) has two connected components. By our assumption, G/N has a normal Sylow p1-subgroup, and so G/N is either Example (2.1) or Example (2.6) from Theorem 2.1. Because p1 is in the smaller component, the group G/N must be Example (2.1), and the components of Δ(G/N) are {p1} and ρ3ρ4. In Example (2.1), G has an abelian Hall p1-complement. Thus for pρ3ρ4, the subgroup Op(G) is a proper subgroup of G that contains P, which is a contradiction to the hypothesis that either Oρ3(G)=G or Oρ4(G)=G. Thus, N=1.

Because Op1(G)=1, the Fitting subgroup of G is P. Let H be a p1-complement for G; H acts faithfully on P, and by [14, Lemma 18.1 and the discussion on p. 254], every prime divisor of |H| occurs in ρ(G). Pick a character γIrr(P) with γ(1)>1 and a character χIrr(Gγ). Then p1 divides χ(1) and no prime in ρ3ρ4 divides χ(1). Thus, G/PH has an abelian Hall ρ3ρ4-subgroup by [14, Theorem 12.9]. Let L=O(ρ1ρ2)\{p1}(H) and E/L=𝔽(H/L). Then H/L has an abelian Hall ρ3ρ4-subgroup, which must be E/L by the Hall–Higman Theorem [5, Theorem 3.21]. If E=H, then we have Op3(H)<H for p3ρ3ρ4, and so Op3(G)<G. This is a contradiction to the hypothesis that either Oρ3(G)=G or Oρ4(G)=G, thus, E<H.

Consider the normal subgroup PE. We know {p1}ρ3ρ4ρ(PE). Since PE<G, we have that Δ(PE) must be disconnected and the components are (ρ1ρ2)ρ(PE) and ρ3ρ4. If p2ρ(PE) for any prime p2(ρ1ρ2)\{p1}, then because PE has a normal Sylow p1-subgroup and both components would have size larger than one, we must be in Example (2.6) of Theorem 2.1. However, p1 must be in the larger component, which it is not, and so PE satisfies Example (2.1) of Theorem 2.1. Because PE is Example (2.1) from Theorem 2.1, the components of Δ(PE) are {p1} and ρ3ρ4, the subgroup E is an abelian Hall ρ3ρ4-subgroup, L=1, and P(PE).

Let Q be the Sylow p4-subgroup of E for some prime p4ρ4. Because PE is Example (2.1) from Theorem 2.1, every non-linear irreducible character of P is fully ramified with respect to P/P(E), the subgroup Q acts faithfully on P, fixing every non-linear irreducible character of P. By [14, Theorem 19.3], we have P=[P,Q]. Let λIrr([P,Q]/P) be non-principal. Because P is central in P by Lemma 4.2, the stabilizer of λ in G is PH(λ).

Because Q acts faithfully, Q(λ)<Q, which implies p4 divides |Q:Q(λ)|. Thus p4 divides |H:H(λ)|. Since p2 and p4 are not adjacent in Δ(G) for p2(ρ1ρ2)\{p1}, we have H(λ) contains a Hall (ρ1ρ2)\{p1}-subgroup of H. Further, λ extends to PH(λ) and

cd(PH(λ)λ)=cd(H(λ)).

So by Clifford’s theory, no degree in cd(H(λ)) is divisible by any prime in (ρ1ρ2)\{p1}. By Itô’s Theorem [3, Theorem 12.34], H(λ) contains a unique Hall (ρ1ρ2)\{p1}-subgroup of H, which is abelian. Recall, E=𝔽(H) is abelian and the index |H:E| is only divisible by the primes in (ρ1ρ2)\{p1}. By [3, Theorem 6.15], cd(H) contains only products of primes in (ρ1ρ2)\{p1} and so ρ(H)=(ρ1ρ2)\{p1}. Thus, H(λ) is abelian. The stabilizer of λ in [P,Q]H is [P,Q]H(λ) and λ extends to this stabilizer.

Consider the group [P,Q]H. We have

cd([P,Q]H)=cd([P,Q]HP)cd([P,Q]HP)
=cd([P,Q]H[P,Q])cd([P,Q]HP|[P,Q]P)cd([P,Q]HP).

Observe that

cd([P,Q]Hλ)={|H:H(λ)|aacd(H(λ))}={|H:H(λ)|}.

Thus the primes in ρ3ρ4 are the only possible prime divisors of degrees in

cd([P,Q]HP|[P,Q]P).

By [6, Lemma 1], [P,Q]/P is irreducible under the action of Q and so

P=[P,Q](E).

Since Δ(PQ) has two connected components, every non-linear irreducible character of [P,Q] is fully ramified with respect to [P,Q]/P. Then [P,Q]E is Example (2.1) from Theorem 2.1 and the two connected components are {p1} and ρ3ρ4.

By [7, Lemma 3.1], cd([P,Q]EP) consists of powers of the prime p1. Since |[P,Q]H:[P,Q]E|=|H:E| is only divisible by primes in (ρ1ρ2)\{p1}, the only primes that divide degrees in cd([P,Q]HP) are the primes in ρ1ρ2. Because Q acts faithfully on [P,Q], the prime p4ρ([P,Q]H). As p4 was chosen arbitrarily from ρ3ρ4, the graph Δ([P,Q]H) has two connected components: ρ1ρ2 and ρ3ρ4. Because [P,Q]H has a normal non-abelian Sylow p1-subgroup and both components have more than one vertex, [P,Q]H must satisfy Example (2.6) from Theorem 2.1. However, p1 is in the component ρ1ρ2 and |ρ1ρ2||ρ3ρ4|. This is a contradiction to [7, Lemma 3.6] and Pálfy’s inequality, Theorem 2.3, and so, G has no normal Sylow p1-subgroups. ∎

As the following argument is used frequently, we have made a lemma. This lemma shows how Theorem 2.6 can be applied once the group G has been shown to have no normal non-abelian Sylow p-subgroups and satisfies the condition that |ρ(G/M)|<|ρ(G)| for all proper nontrivial normal subgroups M.

Lemma 5.3

Let G be a solvable group and suppose G has no normal Sylow p-subgroups for pρ(G). Suppose for all minimal normal subgroups M that ρ(G/M)<ρ(G) and Δ(G/M) does not have diameter three. Then Δ(G) does not have diameter three.

Proof.

We write F for the Fitting subgroup of G and Φ(G) for the Frattini subgroup of G. By [14, Lemma 18.1 and the discussion on p. 254] we know that ρ(G)=π(|G:F|). Because F(G)/Φ(G)=F(G/Φ(G)), we see that

ρ(G/Φ(G))=ρ(G)

and so Φ(G)=1. Let M be a proper nontrivial normal subgroup of G and consider Δ(G/M). Since Δ(G/M) is a subgraph of Δ(G), we see that it cannot have diameter three by hypothesis. Thus, Δ(G/M) must either have two connected components or have diameter two. By Theorem 2.6, we see that Δ(G) must also have two connected components or diameter two. ∎

We now have all the tools to show that the graphs in Figure 4 are not the character degree graphs for any solvable group G.

Proof of Theorem 1.

Let be the family of graphs that are isomorphic to the graphs in Figure 4. In particular, the graphs satisfy Pálfy’s Condition, have six vertices, and have diameter three. Let G be a minimal counter-example with respect to |G|, such that Δ(G). The partition of vertices is as follows: |ρ1ρ2|=2, |ρ3|=1 or 2, |ρ4|2.

Working by induction, if M is a proper normal subgroup of G, then Δ(M) is a subgraph of Δ(G). If Δ(M) has diameter three, then either Δ(M), or Δ(M) is a graph that does not exist as a character degree graph by Theorem 1.2, Theorem 1.3, or by Pálfy’s Condition. As G is minimal, Δ(M) cannot have diameter three and so is either disconnected or has diameter at most two.

If G has a normal Sylow p4-subgroup P for some prime p4ρ4, then

ρ(G/P)=ρ(G)\{p4}

by Lemma 2.4. The graph Δ(G/P) has five vertices. Because ρ4 has more than one vertex, Δ(G/P) either has diameter three or is disconnected. Let p2ρ2 and p3ρ3 be adjacent in Δ(G) such that the product p2p3 divides χ(1) for a character χIrr(G). Because p2 and p4 are not adjacent, we have Pkerχ. Thus, χIrr(G/P) and so p2 and p3 are adjacent in Δ(G/P). Because a graph with five vertices and diameter three does not occur as a character degree graph by Theorem 1.3, G does not have a normal Sylow p4-subgroup. Notice that deleting one or more edges incident to the vertex p4 produces a graph that does not satisfy Pálfy’s Condition, and so Op4(G)=G by [10, Lemma 3.1].

By Lemma 4.5, G does not have a normal Sylow p3-subgroup for any prime p3ρ3. Because |ρ1ρ2|=2 and |ρ3|<3, Lemma 4.6 says that

|ρ(G/M)|<|ρ(G)|

for all proper nontrivial normal subgroups M. We see by Lemmas 5.1 and 5.2 that G does not have a normal Sylow p-subgroup for any prime pρ1ρ2. Thus G has no normal non-abelian Sylow subgroups and so by Lemma 5.3, the graph Δ(G) does not have diameter three, a contradiction. ∎

An immediate corollary is the following:

Corollary 5.4

Let G be a solvable group where Δ(G) has diameter three and |ρ(G)|=6. Then |ρ3|=3.

Proof.

We know from Theorem 3.1 that if G is a solvable group where |ρ(G)|=6, then Δ(G) is either the graph from Figure 3 or the graphs from Figure 4. From Theorem 1 we see that Δ(G) must be the graph in Figure 3. We observe that ρ3 must have three vertices. ∎

Using similar methods to show that the two graphs with six vertices do not occur as diameter three character degree graphs of solvable groups, we can also show that ρ3 must have at least three vertices.

Proof of Theorem 2.

We prove this using induction on the size of ρ(G). Let G be a minimal counter-example such that |ρ(G)|>6, the graph Δ(G) has diameter three, and |ρ3|=1 or 2. Because |ρ1ρ2||ρ3ρ4|, the subset ρ4 has at least two vertices. Suppose G has a normal Sylow p4-subgroup P for some prime p4ρ4. Then by Lemma 2.4, the set of vertices for ρ(G/P) is ρ(G)\{p4}. The graph Δ(G/P) is not disconnected. Let p2ρ2 and p3ρ3 be adjacent in Δ(G) such that the product p2p3 divides χ(1) for a character χIrr(G). Because p2 and p4 are not adjacent, we have Pkerχ. Thus, χIrr(G/P) and so p2 and p3 are adjacent in Δ(G/P). Thus, the graph Δ(G/P) is connected and has diameter three. This contradicts the induction hypothesis and so G has no normal Sylow p4-subgroups for any prime p4ρ4. Furthermore, deleting one or more edges incident to the vertex p4 produces a graph that does not satisfy Pálfy’s Condition, and so, Op4(G)=G by [10, Lemma 3.1].

Let M be a proper, nontrivial, normal subgroup of G. The graphs Δ(M) and Δ(G/M) are subgraphs of Δ(G). In particular, ρ(G/M)ρ3 and ρ(M)ρ3 have at most two vertices. Therefore, the graphs Δ(M) and Δ(G/M) cannot have diameter three.

Since |ρ3|2 and |ρ1ρ2|=a is at least 2, |ρ3|<2a-1. Thus by Lemma 4.5, G does not have any normal Sylow p3-subgroups for any prime p3ρ3. By Lemma 4.6, if M is a normal subgroup of G and ρ(G/M)=ρ(G), then M=1. We see by Lemmas 5.1 and 5.2 that G does not have a normal Sylow p-subgroup for any prime in ρ1ρ2. Hence, G has no normal non-abelian Sylow subgroups. Thus, by Lemma 5.3, Δ(G) does not have diameter three and |ρ3|3. ∎

6 No normal non-abelian Sylow 𝐩𝟒-subgroup

When showing that the graphs in Figure 4 were not the character degree graphs for any solvable group G, we were able to use the fact that the graphs in Figure 2 are not character degree graphs for any solvable group G. Is it possible that there is a group G that has a character degree graph Δ(G) as in Figure 3 and a normal Sylow p4-subgroup P for the prime p4ρ4? The graph Δ(G/P) will have diameter two in this case, and so none of our current tools answer this question. The following lemma shows that it is not possible for a solvable group G to have a character degree graph with diameter three and a normal Sylow p4-subgroup for prime p4ρ4.

Lemma 6.1

Let G be a solvable group with Δ(G) having diameter three. Then G does not have a normal Sylow p4-subgroup for any prime p4ρ4.

Proof.

Let G be a minimal counter-example where Δ(G) has diameter three and G has a normal Sylow p4-subgroup P for a prime p4ρ4. We know that |ρ3|3 by Theorem 2.

Claim 1

We have Op3(G)=G for all primes p3ρ3ρ4\{p4}.

Proof.

Suppose Op3(G)<G. Since Δ(Op3(G)) contains all edges not incident to {p3} and the primes except possibly the prime p3, we can find a prime qρ3 other than p3 that is adjacent to some prime p2ρ2. Thus Δ(Op3(G)) is connected and has diameter three. This contradicts our assumption that G is a minimal counter-example, as POp3(G). ∎

Claim 2

The subgroup Op4(G)=1 and the Fitting subgroup of G is P.

Proof.

Suppose that M is a minimal normal subgroup contained in Op4(G). Then ρ(G/M) contains ρ1ρ2{p4} by Lemma 4.3, and Δ(G/M) is disconnected or has diameter three. Since G/M has a normal Sylow p4-subgroup PM/M, we know that Δ(G/M) must be disconnected by our assumption, and further G/M is either Example (2.6) or Example (2.1) from Theorem 2.1. The components are either ρ1ρ2 and {p4}, or ρ1ρ2 and (ρ3ρ4)ρ(G/M), where |ρ1ρ2|=n and |(ρ3ρ4)ρ(G/M)|2n-1, by Pálfy’s inequality.

We assume first that both of the connected components of Δ(G/M) have size larger than one and so the group G/M is Example (2.6) in Theorem 2.1. Let F/M=𝔽(G/M) and

(E/M)/(F/M)=𝔽((G/M)/(F/M))=E/F=𝔽(G/F).

Then by [7, Lemma 3.6], F/M=PM/M×Z/M where Z/M is a central subgroup of G/M. Let ϕ be a nonlinear irreducible character of the subgroup P so that p4 divides ϕ(1). Then ϕ×1MIrr(P×M) and M is contained in the kernel of ϕ×1M, ie (ϕ×1M)(a)=1 for all aM. By [7, Lemma 3.6], there exists a subgroup B of G such that PM/MB/M is a Frobenius group, PM/M is the Frobenius kernel and B/M is a Frobenius complement. By [3, Theorem 6.34], (ϕ×1M)PB/M is an irreducible character of PB/M, and so (ϕ×1M)PB is an irreducible character of PB.

Let χIrr(G) with χ(1) divisible by p2p3 for primes p2ρ2 and p3ρ3. Let ψIrr(PZ) and γIrr(Z) where ψ is an irreducible constituent of χPZ and γ is an irreducible constituent of ψZ. Notice that if γ=1, then Zker(χ) and so M is also in ker(χ). But then p2 and p3 are adjacent in Δ(G/M), which is not the case, and so γ is not the principal character.

Consider ϕ×γ, an irreducible character of PZ, and its stabilizer Gϕ×γ, which is the intersection of the stabilizers, Gϕ and Gγ. Notice, that by [3, Corollary 6.28], ϕ extends to Gϕ. If θIrr(G) lies over ϕ, then since p4 divides θ(1), we know that no prime in ρ1ρ2 divides θ(1). By [14, Theorem 12.9], we see that G/P contains an abelian Hall ρ1ρ2-subgroup of G/P. As ϕ extends to Gϕ, we can apply Gallagher’s Theorem [3, Corollary 6.17]. Any character degree acd(Gϕ/P) can be multiplied by ϕ(1) to get a character degree in cd(G), and so no prime in ρ1ρ2 divides any character degree of Gϕ/P. Thus, by Itô’s Theorem [3, Corollary 12.34], Gϕ/P contains a unique Hall ρ1ρ2-subgroup AP/P. By Clifford’s theory, we get that no prime in ρ1ρ2 divides |G:Gϕ|.

Further, no prime in ρ1ρ2 divides |G:Gγ×ϕ|. If a prime p1ρ1ρ2 did divide |G:Gγ×ϕ|, then p1 would divide |G:Gϕ| and p1 and p4 would be adjacent, which is not possible. So. since Gϕ×γ/P also contains a Hall ρ1ρ2-subgroup of G, it must contain AP/P. Thus, Gγ contains A.

Consider the character ϕg and its stabilizer Gϕg for some gG. The character ϕg also extends to its stabilizer by [3, Corollary 6.28]. By Gallagher’s theorem [3, Corollary 6.17], any character degree in cd(Gϕg/P) is a multiple of ϕg(1), and so no prime in ρ1ρ2 divides any character degree of GϕG/P. So, this factor group contains a unique Hall ρ1ρ2-subgroup AgP/P. Because Gϕg×γ/P contains Ag/P, Gγ also contains Ag, and so Gγ contains all conjugates of A. Hence, Gγ contains O(ρ1ρ2)(G). However, Gγ contains P and Op3(G)=G for all p3ρ3ρ4\{p4}, and so Gγ=G.

Because 1P×γIrr(P×Z), and E/F is cyclic, we get that 1P×γ extends to E by [3, Corollary 11.22]. Then, as G/E is cyclic, 1P×γ extends to all of the Sylow subgroups of G/E. By [3, Corollary 11.31], we get that 1P×γ extends to G, and hence γ extends to G. Call the extension of γ to Gγ^. By Gallagher’s Theorem, ηγ^ are all of the irreducible constituents of γG, where ηIrr(G/Z). In particular, χ=ηγ^ for some ηIrr(G/Z). Since M is contained in Z, we know that Δ(G/Z) is disconnected and any p2ρ2 and p3ρ3 are not adjacent. Since p2p3 divides χ(1) and p2 does not divide γ(1), we know that p2 must divide η(1). Thus, p3 cannot divide η(1) as p2 and p3 are not adjacent in Δ(G/Z), and so p3 divides γ(1). However, there exists an αIrr(G/Z) such that p1 divides α(1) for some prime p1ρ1 and αγIrr(G). This contradicts the fact that p1 and p3 are not adjacent in Δ(G).

Thus, (ρ3ρ4)ρ(G/M)={p4} and so G/M is Example (2.1) from Theorem 2.1. Then, G/M has an abelian Hall p4-complement, and hence Op3(G)<G for some prime p3ρ3ρ4\{p4}, which we showed cannot happen. Thus we have Op4(G)=1 and 𝔽(G)=P. ∎

Claim 3

The subgroup P is a minimal normal subgroup of G and P is central in P.

Proof.

Suppose there exists 1XP where P/X is a chief factor. Then since ρ(G/P)=ρ(G)\{p4} and Δ(G/P) is connected, we have that ρ(G/X)=ρ(G) and in particular Δ(G/X) has diameter three. However, this contradicts our assumption, and P is minimal normal. Because (P) is characteristic in P and so normal in G, it follows that P(P) is normal in G. Because P is nilpotent, 1<P(P), and so P(P) as P is a minimal normal subgroup of G. ∎

Let θIrr(P) be nonlinear and H a p4-complement in G. Notice that H acts faithfully on P and every prime divisor of |H| occurs in ρ(G). Let χIrr(G) be an irreducible constituent of θG. Then, since no prime in ρ1ρ2 divides χ(1), by [14, Theorem 12.9], G/PH has an abelian Hall ρ1ρ2-subgroup. Let L=Oρ3ρ4\{p4}(H) and E/L=𝔽(H/L). Note that E/L is a ρ1ρ2-subgroup. Since H/L has an abelian Hall ρ1ρ2-subgroup, it follows that E/L is the Hall ρ1ρ2-subgroup by the Hall–Higman Theorem in [5, Theorem 3.21] and since Oρ3ρ4\{p4}(G)=G, E=H. In particular, L is a Hall ρ3ρ4\{p4}-subgroup of H.

Let K be a p2-complement in H for a fixed prime p2ρ2. Note that LK. Since E/L=H/L is abelian, K is normal in H. Because G is a minimal counter-example and PK has a normal non-abelian Sylow p4-subgroup, the graph Δ(PK) is disconnected. The subgroup PK is a Hall p2-complement of G and so we have ρ(PK)=ρ(G){p2}. Also, the only edges in Δ(PK) that are not retained from Δ(G) are those that are incident to the prime p2. The subsets ρ1ρ2\{p2} and ρ3ρ4 induce complete subgraphs of Δ(PK) and so the two connected components of Δ(PK) are (ρ1ρ2)\{p2} and ρ3ρ4. As the prime p4 is in a component with size larger than one, the group PK must be Example (2.6) in Theorem 2.1. Thus, PK=TD where D is an abelian group acting coprimely on the group T. By [7, Lemma 3.6], we have that PT, and T has a p4-complement Q. So, PK=PQD and the primes that divide Q are precisely the primes in ρ3ρ4\{p4}. Thus, Q=L and, as Q is abelian, L is abelian.

Because the group PK=PLD is Example (2.6), the subgroup [PL,D] is a Frobenius group where [P,L] is the Frobenius kernel by [7, Lemma 3.6]. A p4-complement in [PL,D] is [L,D] and so we call a Frobenius complement B=[L,D], which is contained in L. From [7, Lemma 3.6], we have

P=[P,L][P,L].

We see that [P,L]K satisfies the hypotheses of Example (2.6) of Theorem 2.1. So Δ([P,L]K) is disconnected with components (ρ1ρ2)\{p2} and ρ3ρ4. The action of B on P is a Frobenius action, so P(B)=1 and P(L)P(B). Also, P(L)P(L)P=1. Thus, P(L) is abelian.

Let the character λIrr(P) be non-principal. Because P is central in P, the stabilizer of λ is PH(λ). By [3, Theorem 13.28], we can find a H(λ)-invariant irreducible constituent θ of λP. Note that the stabilizer of θ in G is PH(θ) and H(λ)H(θ). Because P is central and θP has a unique constituent λ, we have that H(λ)=H(θ).

As p4 divides every degree in cd(Gλ), we have that H(λ) contains an abelian Hall ρ1ρ2-subgroup of H. Further, θ must extend to PH(λ) by [3, Corollary 6.28] and so, by Gallagher’s Theorem, no prime in ρ1ρ2 is in ρ(H(λ)). Since λ extends to PH(λ), and p4 divides every degree in cd(Gλ), we see that no prime in ρ1ρ2 divides any degree in cd(PHP). On the other hand, H has a normal abelian Hall ρ3ρ4\{p4}-subgroup L, and so no prime in ρ3ρ4 divides a degree in cd(H). If H is nilpotent, then by [14, discussion on p. 254], there would be a character degree that equals |H|. Since all of the primes in ρ(G)\{p4} divide |H|, this is not the case. Because L𝔽(H), we deduce that p1ρ(H) for some p1ρ1ρ2\{p2}. Recall that L contains B, so that PB is a Frobenius group. It follows that at least one prime in ρ3ρ4\{p4} is in ρ(PH) and Δ(PH) is disconnected. The components are a nonempty subset of ρ1ρ2 and a nonempty subset of ρ3ρ4\{p4}. Because PH has no normal non-abelian Sylow subgroups, PH is Example (2.4) from Theorem 2.1.

First, we suppose that ρ1ρ2ρ(PH). Recall that PK is Example (2.6) from Theorem 2.1 and so by [7, Lemma 3.6], PK/P is Example (2.4). Because L is a normal Hall subgroup of H, we know L𝔽(K). By [7, Lemma 3.6], only the primes in ρ3ρ4 can divide 𝔽(K). Thus, 𝔽(K)L and L=𝔽(K). Further, both L and K/L are cyclic groups.

Consider the subgroup HP/P acting coprimely on the group [P,L]/P. Define E/P=𝔽(HP/P)𝔽(H). Certainly, L𝔽(H). Because p2 divides |H| and K is a normal p2-complement of H, the Fitting subgroup of K is contained in the Fitting subgroup of H. Hence, 𝔽(H)=L×Op2(H). Because the Sylow p2-subgroup of H is not a normal subgroup of H, the subgroup Op2(H) is a proper subgroup of H. Further, the subgroup PKOp2(H) is a proper normal subgroup of G and ρ(PKOp2(H))=ρ(G). As the subgroup PKOp2(H) has a normal Sylow p4-subgroup, and by the minimality of G, it follows that the graph Δ(PKOp2(H)) must be disconnected. Both components have size larger than one and so PKOp2(H) is Example (2.6) from Theorem 2.1. By [7, Lemma 3.6], the primes that divide |𝔽(H):𝔽(PKOp2(H))| are precisely the primes in ρ3ρ4, and so p2 cannot divide that index, which is a contradiction. Hence, Oρ2(H)=1 and 𝔽(H)=L and E/P=LP/P.

Recall that the group [P,L]K is Example (2.6) from Theorem 2.1. By [7, Lemma 3.6], the factor group [P,L]K/P is Example (2.4) from Theorem 2.1. First, we define

Z1/P=KP/P([P,L]/P)andZ2/P=HP/P([P,L]/P).

As 𝔽(G/P)=P/P and G/P(P/P)P/P, we see that

Z1/P=Z2/P=P/P.

The Fitting subgroup of K is L and so the Fitting subgroup of KP/P is LP/P and it is abelian. As [P,L]K is Example (2.6) from Theorem 2.1, L acts irreducibly on [P,L], we have LP/P acts irreducibly on [P,L]/P. Define

m=|HP/P:LP/P|,

which is equal to |H:L|, and so (m,|L|)=1. Let m2=|H:K| be the power of p2 that divides |H|.

Let q1 be a power of p4 such that q1m/m2=|P/P|. Because [P,L]K/P is Example (2.4), there is a p4-power q such that qm/m2=|[P,L]:P|. Clearly, we have qq1. Let s be a Zsigmondy prime divisor of q1m/m2-1. We recall that Zsigmondy prime divisors exist except if m=2 and p4=2 or m=6 and q=2. These exceptions do no occur as [P,L]K/P is Example (2.4). Then s divides q1m/m2-1 and does not divide q1i-1 for any i<m/m2. Further, as q1-1 is a factor of q1m/m2-1, the prime s divides (q1m/m2-1)/(q1-1), and as the quotient (q1m/m2-1)/(q1-1) divides |L|, so does s. Now the subgroup L acts Frobeniusly on [P,L]/P. So L divides |[P,L]/P|-1=qm/m2-1. Hence, s divides qm/m2-1, and since s is a Zsigmondy prime divisor, we have qm/m2q1m/m2. Thus q=q1.

Because PH is also Example (2.4), we know that there is a p4-power q2 such that q2m=|P|. Because L acts Frobeniusly on P, we know that |L| divides q2m-1. Because s divides |L|, we know that s divides q2m-1 and so qm/m2q2m. Let r be a Zsigmondy prime divisor of q2m-1. The prime r exists because the exceptions do not occur as PH is Example (2.4). Then r divides q2m-1 and r also divides (q2m-1)/(q2-1). Hence, r divides |L|. But as |L| divides qm/m2-1, so does r. But then q2mqm/m2 and so q2m=qm/m2. Hence, q2=q1/m2. Since (q2m-1)/(q2-1) divides |L|, we know that (qm/m2-1)/(q1/m2-1) divides |L|. Thus [P,L]H/P satisfies Example (2.4) and so the graph Δ([P,L]H/P) is disconnected.

By Fitting’s Lemma, P/P=P/P(L)×[P,L]/P. So

PH/P=P/PHP/P=(P/P(L)×[P,L]/P)HP/P.

If P(L)/P is not central in HP/P, then there is a character degree divisible by |H:H(λ)|θ(1). Because K is a p2-complement, p2 divides |H:H(λ)|, and so p2 and p4 are adjacent. This is a contradiction, and so ρ2 is not contained in ρ(PH). In particular, the prime p2 is not in ρ(H). So H has an abelian Sylow p2-subgroup Q by [3, Corollary 12.34]. Further, PQ is a normal subgroup of G.

Now, the graph Δ(PQ) has two connected components, {p2} and {p4}. Observe that Q acts coprimely on P, fixing all non-linear characters. From [14, Theorem 19.3], we have [P,Q]=P and [P,Q] is not abelian. Consider Δ([P,L]H) determined by

cd([P,L]H)=cd([P,L]H[P,L])cd([P,L]H[P,L]).

Because ρ(H)=(ρ1ρ2)\{p2}, the prime p2 does not divide a character degree in cd(H)=cd([P,L]H/[P,L]). Recall that [P,L]K is Example (2.6) from Theorem 2.1 and B is a Frobenius complement. Thus by [7, Lemma 3.6], every degree in cd([P,L]K[P,L]) is divisible by p4|B| and so p2 does not divide any of those character degrees. We see that p2 is not in cd([P,L]H), and so the subgroup [P,L]H is proper in G. The graph Δ([P,L]H) must be disconnected. Thus, Q is a normal subgroup of [P,L]H. In particular, Q centralizes [P,L] as Q is normal in H. Because [[P,L],Q]=1 and [L,Q]=1, we have [[L,Q],P]=1. Thus [[Q,P],L]=1 by the Three Subgroup Lemma [5, Lemma 4.9], and so [P,Q]P(L). This is a contradiction because P(L) is abelian and [P,Q] is not. ∎

7 Main theorems

We now prove that when G is a solvable group with Δ(G) having diameter three, then G must have exactly one normal non-abelian Sylow p-subgroup and pρ3. Because Theorem 2.5 tells us that G has at most one normal non-abelian Sylow p-subgroup when G is solvable and Δ(G) has diameter three, it is possible that a solvable group G could have no normal non-abelian Sylow p-subgroups and have a character degree graph Δ(G) that has diameter three.

In our proof, we assume that G has no normal non-abelian Sylow p-subgroups for pρ3 and our goal is to use Lemma 5.3 to show that Δ(G) could not have had diameter three. To do this, we must show that G has no normal non-abelian Sylow p-subgroups for any prime p. We will apply Lemma 6.1 to show that G has no normal non-abelian Sylow p-subgroups for any prime pρ4. To show that there are no normal non-abelian Sylow p-subgroups for any prime pρ1ρ2 we must apply Lemma 5.2 and Lemma 5.1. Most of the work of this proof is to show that the hypotheses of these lemmas are met.

Theorem 3

Let G be a solvable group with character degree graph Δ(G) with diameter three. Then G has a normal Sylow p-subgroup for exactly one prime p and pρ3.

Proof.

Let G be a counter-example with |G| minimal such that Δ(G) has diameter three and G has no normal Sylow p3-subgroups for p3ρ3. Because Δ(G) has diameter three, we know |ρ3|3 by Theorem 2. By Lemma 6.1, we see that G does not have a normal Sylow p4-subgroup for any prime p4ρ4.

Claim 1

Let p2 be a prime in ρ2. If Op2(G) is a proper subgroup of G, then Δ(Op2(G)) is disconnected.

Proof.

Let K=Op2(G) for some prime p2ρ2. Since ρ(K) contains ρ1ρ4, the graph Δ(K) either has diameter three or is disconnected. Suppose Δ(Op2(G)) has diameter three. Then by the hypothesis, Op2(G) has a normal Sylow p3-subgroup P for some prime p3ρ3. The subgroup P is characteristic in Op2(G) and so is normal in G, which is a contradiction. Hence, Δ(Op2(G)) cannot have diameter three. Because ρ(Op2(G)) contains all the primes of ρ(G) with the possible exception of the prime p2, we see Δ(Op2(G)) must be disconnected. ∎

Claim 2

The graph Δ(G/N) cannot have diameter three for any proper, nontrivial, normal subgroup N.

Proof.

Suppose there exists a normal subgroup N of G such that Δ(G/N) has diameter three. Then we can find a minimal normal subgroup M contained in N, where Δ(G/M) has diameter three. By the minimality of G, G/M has a normal Sylow p3-subgroup P/M for p3ρ3. Because M is an elementary abelian p-group for some prime p, if p=p3, then G has a normal Sylow p3-subgroup and so, pp3.

Because P1, either MP or MP=1. Suppose that MP=1. Then because [P,M]P and [P,M]M, we see [P,M]=1 and so M is central in P. Let P3 be a Sylow p3-subgroup of G such that P3P. Since M normalizes P3, P3 is characteristic in P and so is normal in G. This is a contradiction, and so MP. Hence G/M/(P/M)=G/M/P/MG/P.

Suppose that G has a normal Sylow q-subgroup Q for qρ1ρ2ρ4. Then we have a contradiction of Lemma 4.4 because G/M has a normal Sylow p3-subgroup. Hence, G has no normal non-abelian Sylow subgroups.

Consider the graph Δ(G/M/(P/M))=Δ(G/P). By our hypothesis, if the graph Δ(G/P) has diameter three, then it has a normal Sylow q-subgroup for qρ3. Since qp3, this contradicts the fact that G/M can have at most one normal non-abelian Sylow p-subgroup. Hence, G/P does not have diameter three and further, does not have any normal Sylow p-subgroups for pρ(G/P). Because ρ(G/P) contains primes from ρ1 and ρ4, we know that Δ(G/P) must be disconnected and is Example (2.4) from Theorem 2.1.

Let p2 be a prime in ρ2. By [7, Theorem 5.3], Op2(G/P)<G/P and so Op2(G)<G. Recall that G has no normal non-abelian Sylow p-subgroups for pρ(G). Suppose Op2(G) has a normal non-abelian Sylow t-subgroup T. Then because T is normal in Op2(G) and Op2(G) is characteristic in G, the subgroup T is normal in G. This is a contradiction, and so Op2(G) has no normal non-abelian Sylow p-subgroups. However, Δ(Op2(G)) is disconnected and so is either Example (2.4) or Example (2.5) from Theorem 2.1 which contradicts Lemma 4.1. Thus, Δ(G/N) cannot have diameter three for any proper, nontrivial, normal subgroup N. ∎

Claim 3

The graph Δ(M) does not have diameter three for any proper normal subgroup M of G.

Proof.

Suppose that M is a proper normal subgroup of G such that Δ(M) has diameter three. Then M has a normal Sylow p3-subgroup P for a prime p3ρ3. Since P is characteristic in M, and M is normal in G, we see P is normal in G. Further, P is normal in G, and P1. Recall that Δ(G/P) cannot have diameter three. Because ρ(G/P) contains ρ(G)\{p3}, the graph has components ρ1ρ2 and ρ3ρ4\{p3}, with the possibility of containing p3 as well. It must be disconnected with both components at least size 2. By [7, Theorem 5.3], we have Op2(G/P)<G/P for some prime p2ρ2, and Op2(G)<G. Further, POp2(G). We have shown that Δ(Op2(G)) is disconnected and Lemma 4.1 tells us that Op2(G) cannot be Example (2.4) or Example (2.5) from Theorem 2.1.

Suppose |ρ1ρ2|=2 and p2ρ(Op2(G)). Then it is possible that Op2(G) is Example (2.1) from Theorem 2.1 and Op2(G) has a normal Sylow p1-subgroup R for p1ρ1. Since Δ(M) has diameter three, p1ρ(M) and so M has a normal Sylow p1-subgroup, which contradicts the fact that M can have at most one normal non-abelian Sylow p-subgroup. Thus Op2(G) cannot be Example (2.1) from Theorem 2.1. Since at least one component, if not both, has size at least 2, Op2(G) is Example (2.6) from Theorem 2.1. Hence Op2(G) has a normal Sylow p-subgroup Q for a prime pρ3ρ4. But this is a contradiction because Q is characteristic in Op2(G) and so Q is a normal Sylow p-subgroup of G and G does not have a normal Sylow p-subgroup for any prime in ρ3ρ4. Thus, Δ(M) cannot have diameter three whenever M is a proper normal subgroup of G. ∎

Claim 4

The subgroup Op3(G)=G for all primes p3ρ3.

Proof.

Suppose there exists a prime p3ρ3 such that Op3(G) is proper in G. Since Op3(G) is proper in G we know that Δ(Op3(G)) cannot have diameter three. Because ρ(Op3(G)) contains all of ρ(G) except perhaps the prime p3, we know that Δ(Op3(G)) is disconnected. Because |ρ3|3, there is a prime qρ3, not equal to p3, and a prime p2ρ2, such that q and p2 are adjacent in Δ(G). However, as Δ(Op3(G)) contains all edges not incident to the prime p3, we see that p2 and q are adjacent in Δ(Op3(G)) and Δ(Op3(G)) must have diameter three. This is a contradiction. ∎

Claim 5

If M is a normal subgroup of G, then ρ(G/M)=ρ(G) implies that M=1.

Proof.

Suppose there exists a nontrivial normal subgroup M of G such that ρ(G/M)=ρ(G). Without loss of generality, M is a minimal normal subgroup of G. We know that the graph Δ(G/M) cannot have diameter three, and because ρ(G/M) contains primes from ρ1 and ρ4, the graph Δ(G/M) must be disconnected. The components are ρ1ρ2 and ρ3ρ4. Because ρ1ρ2 is the smaller component and has size at least 2, G/M is either Example (2.4) or Example (2.6) from Theorem 2.1. Suppose G/M is Example (2.6). Then G/M has a normal Sylow p-subgroup P/M for a prime pρ3ρ4. Let Q be a Sylow p-subgroup of G such that Q is contained in P. If M is not contained in P, then since [P,M]P and [P,M]M, we have [P,M]=1. Thus M is central in P and, as M normalizes Q, we have Q is characteristic in P and so Q is normal in G, which is a contradiction and so MP. Since ρ(G/M)/(P/M))=ρ(G)\{p} and G/M/(P/M)=G/P, the graph Δ(G/P) is disconnected. The factor group G/P is Example (2.4) from Theorem 2.1. By [7, Lemma 5.3], Op2(G/P) is proper in G/P and so Op2(G)<G. But this contradicts Lemma 4.1, thus, G/M is not Example (2.6), and so is Example (2.4). If G has a normal non-abelian Sylow subgroup, then so does G/M. Therefore, G has no normal non-abelian Sylow subgroups. By [7, Theorem 5.3], Op2(G/M)<G/M. But then Op2(G)<G, which contradicts Lemma 4.1. Thus, ρ(G/M)=ρ(G) implies that M=1. ∎

By Claim 2, Claim 3, Claim 4, and Claim 5, the hypotheses for Lemma 5.1 and Lemma 5.2 are satisfied. Thus, G does not have a normal non-abelian Sylow p-subgroup for any prime pρ1ρ2. Recall that G has no normal non-abelian Sylow p-subgroup for pρ4 and so G has no normal non-abelian Sylow p-subgroups. Thus, by Lemma 5.3, G does not have diameter three, which is our final contradiction. ∎

Finally, because G has a normal Sylow p3-subgroup P when Δ(G) has diameter three, we can observe that Δ(G/P) must be disconnected, and so, G/P is in one of the families of groups from Theorem 2.1.

Theorem 4

Let G be a solvable group with Δ(G) having diameter three. If |ρ1ρ2|=n, then |ρ3ρ4|2n

Proof.

By Theorem 3, G has a normal non-abelian Sylow p3-subgroup P for some prime p3ρ3. By Lemma 2.4, ρ(G/P)=ρ(G)\{p3}. Because ρ3 has more than three vertices by Theorem 2, either Δ(G/P) is disconnected or it has diameter three. If Δ(G/P) has diameter three, then by Theorem 3, Δ(G/P) has a normal non-abelian Sylow q-subgroup Q/P for some prime qρ(G)\{p3}. Let R be a Sylow q-subgroup of G contained in Q. As RP/P is a normal subgroup of G/P, the Sylow subgroup R is a normal subgroup of G, which contradicts the fact that G can have at most one normal non-abelian Sylow p-subgroup. Hence, Δ(G/P) is disconnected. By Pálfy’s inequality, if |ρ1ρ2|=n, then |ρ3ρ4\{p3}|2n-1. Hence, |ρ3ρ3|2n. ∎

Theorem 5

Let G be a solvable group with Δ(G) having diameter three. Then G has Fitting height 3.

Proof.

By Theorem 3, G has a normal Sylow p3-subgroup P for p3ρ3. So G/P has a normal abelian Sylow p3-subgroup. Notice that Δ(G/P) is disconnected, both components have size larger than 2, and G/P has no normal Sylow p-subgroups for pρ(G/P). So G/P is Example (2.4) from Theorem 2.1 and so has Fitting height 3. Let H be a p3-complement of G. Anything in H that centralizes P/P also centralizes P. Let F=𝔽(G) and E/P=𝔽(G/P). We have F/P is a nilpotent normal subgroup of G/P and FE. Conversely, E=P(EH), and EH is a nilpotent subgroup that centralizes P/P. So E=P×(EH) is nilpotent and EF. Thus the Fitting height of G is the same as G/P, and so G has Fitting height three. ∎


Communicated by Robert M. Guralnick


Acknowledgements

This work was completed under the direction of Mark Lewis. and is included in the writer’s PhD dissertation. It was done independently of the work of Carlo Casolo, Silvio Dolfi, Emanuele Pacifici, and Lucia Sanus in [1]. While the main results in Theorems 4 and 5 are shown in both papers, the methods are different. I would like to thank Dr. Gagola, Dr. White, and especially Dr. Lewis for all of their help and support during the writing of this.

References

[1] Casolo C., Dolfi S., Pacifici E. and Sanus L., Groups whose character degree graph has diameter three, preprint. 10.1007/s11856-016-1387-5Search in Google Scholar

[2] Dugan C. T., Solvable groups whose character degree graphs have diameter three, Ph.D. thesis, Kent State University, 2007. Search in Google Scholar

[3] Isaacs I. M., Character Theory of Finite Groups, Academic Press, New York, 1976. Search in Google Scholar

[4] Isaacs I. M., Algebra. A Graduate Course, American Mathematical Society, Providence, 1994. Search in Google Scholar

[5] Isaacs I. M., Finite Group Theory, American Mathematical Society, Providence, 2008. 10.1090/gsm/092Search in Google Scholar

[6] Lewis M. L., Bounding Fitting heights of character degree graphs, J. Algebra 242 (2001), no. 2, 810–818. 10.1006/jabr.2001.8831Search in Google Scholar

[7] Lewis M. L., Solvable groups whose degree graphs have two connected components, J. Group Theory 4 (2001), no. 3, 255–275. 10.1515/jgth.2001.023Search in Google Scholar

[8] Lewis M. L., A solvable group whose character degree graph has diameter 3, Proc. Amer. Math. Soc. 130 (2002), no. 3, 625–630. 10.1090/S0002-9939-01-06091-9Search in Google Scholar

[9] Lewis M. L., Solvable groups with degree graphs having 5 vertices and diameter 3, Comm. Algebra 30 (2002), no. 11, 5485–5503. 10.1081/AGB-120015664Search in Google Scholar

[10] Lewis M. L., Classifying character degree graphs with 5 vertices, Finite Groups 2003 (Gainesville 2003), Walter de Gruyter, Berlin (2004), 247–265. 10.1515/9783110198126.247Search in Google Scholar

[11] Lewis M. L., An overview of graphs associated with character degrees and conjugacy class sizes in finite groups, Rocky Mountain J. Math. 38 (2008), no. 1, 175–210. 10.1216/RMJ-2008-38-1-175Search in Google Scholar

[12] Manz O., Degree problems II: π-separable character degrees, Comm. Algebra 13 (1985), no. 11, 2421–2431. 10.1080/00927878508823281Search in Google Scholar

[13] Manz O., Willems W. and Wolf T. R., The diameter of the character degree graph, J. Reine Angew. Math. 402 (1989), 181–198. 10.1515/crll.1989.402.181Search in Google Scholar

[14] Manz O. and Wolf T. R., Representations of Solvable Groups, Cambridge University Press, Cambridge, 1993. 10.1017/CBO9780511525971Search in Google Scholar

[15] Pálfy P. P., On the character degree graph of solvable groups. I: Three primes, Period. Math. Hungar. 36 (1998), no. 1, 61–65. 10.1023/A:1004659919371Search in Google Scholar

[16] Pálfy P. P., On the character degree graph of solvable groups. II: Disconnected graphs, Studia Sci. Math. Hungar. 38 (2001), 339–355. 10.1556/sscmath.38.2001.1-4.25Search in Google Scholar

[17] Zhang J., On a problem by Huppert, Beijing Daxue Xuebao Ziran Kexue Ban 34 (1998), no. 2–3, 143–150. Search in Google Scholar

[18] Zuccari C. P. M., Fitting height and diameter of character degree graphs, Comm. Algebra 41 (2013), no. 8, 2869–2878. 10.1080/00927872.2012.665534Search in Google Scholar

Received: 2014-5-13
Revised: 2014-11-21
Published Online: 2016-7-9
Published in Print: 2016-11-1

© 2016 by De Gruyter

Downloaded on 29.10.2025 from https://www.degruyterbrill.com/document/doi/10.1515/jgth-2016-0029/html
Scroll to top button