Group Orders That Imply a Nontrivial p-Core

Let G be a group of order p^{r}m with O_{p}(G)=1. Let P be a Sylow p-subgroup of G. Since the kernel of the action of G on the set of cosets of P is precisely O_{p}(G), we obtain that G embeds in S_{m}, and so p^{r} divides (m-1)!. Hence, if p^{r_{0}} is the largest power of p dividing (m-1)!, we obtain that \Lambda(p,m)\leq r_{0}+1.

Let K be the group C_{q_1}^{t_1}\times\cdots\times C_{q_k}^{t_k}, that is, a product of elementary abelian groups, where m=q_{1}^{t_1}\cdots q_{k}^{t_k} and q_{1},\ldots,q_{k} are distinct primes and C_{q} denotes the cyclic group of order q. Then \Gamma(m) divides the order of \mathop{\mathrm{Aut}}(K), and hence so does p^{s}. Let H be a subgroup of \mathrm{Aut}(K) of order p^{s}. For every S\in H and k\in K define the map T_{S,k}\colon K\to K by T_{S,k}(x)=Sx+k. Then G=\left\{\,T_{S,k}\mid S\in H, k\in K\,\right\} is also a subgroup of \mathrm{Aut}(K). If we identify H with the subgroup of maps of the form T_{S,0} and K with the subgroup of maps of the form T_{1_K,k}, then G is just the semidirect product of K by H. Hence |G|=n. We have that G acts transitively on K in a natural fashion, and the stabilizer of 0\in K is H, a p-Sylow subgroup of G. Hence the stabilizers of points in K are precisely the Sylow subgroups of G, so their intersection O_{p}(G) contains only the identity K\to K, as we wanted to prove.

Let G be solvable with order n=p^{s}m and O_{p}(G)=1. Let F(G) be the Fitting subgroup of G. Consider the map c\colon G\to\mathop{\mathrm{Aut}}(F(G)), sending g to c_{g}\colon F(G)\to F(G) given by conjugation by g. The restriction of c to P, a p-Sylow subgroup of G, has kernel P\cap C_{G}(F(G)). Since C_{G}(F(G))\le F(G) (Theorem 7.67 from cite MR1298629-djvu), and F(G) does not contain elements of order p by our assumption on O_{p}(G), we have P\cap C_{G}(F(G))=1 and so P acts faithfully on F(G). If m=q_{1}^{t_1}\cdots q_{k}^{t_k} is the prime factorization of m, we have that F(G) is the direct product of the O_{q_{i}}(G) for i=1,\ldots,k. Hence P\le\mathop{\mathrm{Aut}}(F(G))\cong \mathop{\mathrm{Aut}}(O_{q_1}(G))\times\cdots\times \mathop{\mathrm{Aut}}(O_{q_k}(G)). Let g\in P such that the action induced by c_{g} on \prod_{i} O_{q_i}(G)/\Phi(O_{q_i}(G)), is the identity. Since c_{g} acts on each factor O_{q_i}(G)/\Phi(O_{q_i}(G)) as the identity, then by Theorem 5.1.4 from cite MR569209-big, we have that it acts as the identity on each O_{q_i}(G). By the faithful action of P on F(G), we have that g=1. This implies that P acts faithfully on \prod_{i} O_{q_i}(G)/\Phi(O_{q_i}(G)). But then |P| divides the order of the automorphism group of \prod_{i} O_{q_i}(G)/\Phi(O_{q_i}(G)), which is a product of elementary abelian groups of respective orders q_{i}^{s_i} with s_{i}\le t_{i} for all i. Hence p^{s}=|P| divides \Gamma(m).

By Burnside’s p,q-theorem, and the Odd Order Theorem, we have that all groups that have order of the form p^{r}m for some r are solvable. Therefore, for all r>s, by Theorem r@solvablesolvable we have that all groups of order p^{r}m have non-trivial p-core.

With the notation of the argument of the proof of r@solvablesolvable, if G is a solvable group of order n with O_{p}(G)=1, we must have that |O_{q_{i}}(G)|=q_{i}^{t_{i}} and \Phi(O_{q_i}(G))=1 for all i in order to satisfy the divisibility conditions. Hence O_{q_{i}}(G) is elementary abelian and a q_{i}-Sylow subgroup for all i, and so G is the semidirect product of a p-Sylow subgroup P of F(G)=C_{q_1}^{t_1}\times\cdots\times C_{q_k}^{t_k} with F(G), where the action of P on F(G) by conjugation is faithful. Hence G is isomorphic to the group constructed in the proof of Theorem r@solvablesolvable.

Let G be a group of order 2^{r}\cdot 3\cdot 5 for r\geq4. Suppose that O_{2}(G)=1. From Theorem r@solvablesolvable, we obtain that G is not solvable. We will prove then that O_{3}(G)= 1. Suppose otherwise, and let T=O_{3}(G). Then |G/T|=2^{r}\cdot 5, and so G/T is solvable. Since 2^{r}\nmid \Gamma(5), from Theorem r@solvablesolvable, we have that O_{2}(G/T)\ne 1. Let L\lhd G such that O_{2}(G/T)=L/T. Suppose |L/T|=2^{j}. Since O_{2}(G/L)=1, |G/L|=2^{r-j}\cdot 5 and G/L is solvable, we have that 2^{r-j} divides \Gamma(5)=2^{2}, that is, r-j\leq 2. Now, L is also solvable and \Gamma(3)=3-1=2, hence if we had j\geq 2 we would have O_{2}(L)\ne 1, and G would have a non-trivial subnormal 2-subgroup, which contradicts our assumption that O_{2}(G)=1. Hence j=1. But then r-1\leq 2, which contradicts that r\geq 4. Hence O_{3}(G)=1. By a similar argument, we get that O_{5}(G)=1.

From cite 1968-brauer-on-simple-groups-of-order-5dot-3-spadot-2-spb we obtain that G is not simple. Hence G has a proper minimal normal subgroup M. From the previous paragraph, we obtain that M is not abelian, since in that case we would have that M\leq F(G). The only possibility is that M=A_{5}. We have then a morphism c\colon G\to \mathrm{Aut}(A_{5}) sending g to c_{g}, the conjugation by g. Since \mathrm{Aut}(A_{5})=S_{5}, and |c(G)|=|\mathrm{Inn}(G)|\geq |\mathrm{Inn}(A_{5})|=60, in any case the kernel of c is a nontrivial normal 2-subgroup.