Wedderburn decomposition of Twisted skew abelian group algebra

As G and \mathbb{F}_{p^n}^* are groups, we have that \hat{G} is closed under the operation. Also, the associativity comes from the associativity of the product in a twisted skew group ring.

It has an identity element, given by 1_{\hat{G}} = (1_{\mathbb{F}_p^n}, 1_G) as for all (x,y) \in \hat{G}

And therefore (x,y) 1_{\hat{G}} = (x,y) = 1_{\hat{G}} as we wanted to prove. Finally, the inverse element of (x,y) \in \hat{G} is given by \left( \left(\frac{1}{x\alpha(y,y^{-1})}\right)^{y^{-1}},y^{-1} \right), as

and the other side is proven analogously.

We will prove that the epimorphism \psi :\mathbb{F}_p \hat{G} \longrightarrow \mathbb{F}_{p^n}^\alpha[G,\sigma] is given by \psi(a \overline{(b,c)}) = ab \overline{c} for all a\in \mathbb{F}_p, (b,c) \in \hat{G} and extended by linearity. It respects addition by definition, so we only have to prove that it respects the product, and to do so we can focus on the basis elements

And therefore, it is a \mathbb{F}_p-homomorphism of rings. To prove that it is an \mathbb{F}_p-epimorphism, note that as an \mathbb{F}_p-algebra, it is generated by \lambda^i\overline{g} for i=0,\cdots, n-1 and g\in G, and clearly \lambda^i\overline{g} = \psi((\lambda^i,g)).

To see that the kernel is given by \langle P((\lambda,1)) \rangle, first note that \psi ( P(\lambda,1)) = P(\lambda) \overline{1} = 0. For the other implication, let \sum_{i=0,\cdots, n; g\in G}a_{i,g} ( \lambda^i, g) \in \mathbb{F}_p \hat{G} such that \psi( \sum_{i=0,\cdots, n; g\in G}a_{i,g} ( \lambda^i, g) ) = 0. Then we have that

So \sum_{i=0}^n a_{i,g} \lambda^i =0 for all g\in G. As p(x) is the minimal polynomial of \lambda, we have that P(x) | \sum_{i=0}^n a_{i,g} x^i and \sum_{i=0}^n a_{i,g} x^i = k_g(x)P(x). As a result,

Let \pi_2 : \hat{G} \longrightarrow G be the projection on the second component. It is trivial to check that it is an epimorphism of groups and that \ker (\pi_2) = \mathbb{F}_{p^n} \times \{1\}. As the 2-cocycle is normalised, and 1_G acts trivially over \mathbb{F}_{p^n}, we have that \mathbb{F}_{p^n} \times \{1\} \simeq \mathbb{F}_{p^n}, and therefore it is commutative. In addition, as it is the kernel of a morphism, it is a normal subgroup. Therefore, \ker(\pi_2) \trianglelefteq \hat{G} is an abelian normal subgroup such that \hat{G} / \ker(\pi_2) \simeq Im(\pi_2) = G is abelian; thus, \hat{G} is metabelian, as we wanted to prove.

As Q_{\hat{H}} , Jin(\hat{H}) \subseteq H we have that Q_{\hat{H}} \cdot Jin(\hat{H}) \leq H. To prove the other inclusion, fix \psi such that Jin(\hat{H}) = Jin_\psi (H). Let (a,b)\in H. If (a,b)\in Jin(\hat{H}) we have finished. If (a,b) \not \in Jin(\hat{H}), then there exist (\psi(b),b) \in Jin(\hat{H}) \leq H and therefore (a,b) (\psi(b),b)^{-1} \in H. Then

So \left(\frac{a}{\psi(b)},1\right) \in Q_{\hat{H}}, and \left(\frac{a}{\psi(b)},1\right) \cdot (\psi(b),b) = (a,b) as we wanted to prove.

First of all, we have that (\psi(a),a), (\psi(b),b) \in H so

As we have that (\psi(ab),ab) \in H, then by the definition of Q_{\hat{H}}, we have that \frac{\psi(ab)}{\psi(a) \psi(b)^a \alpha(a,b)} \in Q_{\hat{H}}. The first results come from induction on the number of products.

For the second part of the proposition, we know that if \psi gives a Jineta of H, then it must satisfy the previous identity. In addition, as G is abelian, we have that \psi(ab) = \psi(ba). Therefore, we have that

Which leads to

As \mathbb{F}_q^* is a cyclic group, Q_{\hat{H}} is the only subgroup of order |Q_{\hat{H}}|, and it has all elements x\in \mathbb{F}_q^* such that x^{|Q_{\hat{H}}|}. For the second identity, we must have that \pi_2((\psi(a),a)^{ord(a)}) = a^{ord(a)} = 1, so (\psi(a),a)^{ord(a)}) \in Q_{\hat{H}}. Therefore, we have that

Reasoning as in the previous case, we have that

Now, for the if part of the second statement, suppose that those two equations are satisfied. Then, all elements in H can be written as a word in terms of C, a=\prod_{i=1}^n a_i and we can define

To check that this map is well-defined, it suffices to show that \psi^*(AabB) = \psi^*(AbaB) for all A, B \in G and a, b \in C. The first identity of equation (8) provide that the word of length 2 are well defined. By induction, lets suppose that the words up to n-1 are well defined. Then, we have that

Therefore, we can define \psi(a) as a representative of \psi^*(a). Finally, it is trivial to see that \{(\psi(a),a) ; a \in \pi_2(H)\} \subset H and therefore it define a Jineta for H

To prove this, let \psi(a) = \lambda^{\Psi(a)}. Then the first equation is the first equation of (8), while for the second one, note that

So our new second equation is the second equation of (8). Therefore, \psi can be extended from C to \langle C \rangle = H, and the set B=Q\cdot \{(\psi(a),a) ; a\in H\} is a group that satisfies our conditions.

If we denote by \pi_H the projection G over G/H, then we have the diagram

where \gamma is an homomorphism from \hat{G} to G/H. As \hat{H} \subseteq\ker(\gamma), then g induces a homomorphism of group \gamma':\hat{G}/\hat{H} \longrightarrow G/H, given by \gamma'((a,b)\hat{H}) = bH. Therefore \pi_2 = \gamma' and we have the desired result.

First, suppose that \hat{G}/\hat{H} is cyclic. Then G/H =\pi'_2(\hat{G}/ \hat{H}) is also cyclic. In addition, if (a,t)\hat{H}, clearly \pi_2'( (a,t) \hat{H}) =t H is some generator of G/H.

Decompose in cyclic components K= \prod_{i,j} C_{p_i^{k_{i,j}}} with p_i prime, k_{i,j}\in \mathbb{N}. Then H=\prod_{i,j} C_{p_i^{h_{i,j}}} with h_{i,j} \leq k_{i,j} and h_{i,j} \not = k_{i,j} for at most one j for each i. Let \lambda=(\lambda_{i.j}) \in K given by \lambda_{i,j}=1 for all i,j such that h_{i,j} = k_{i,j} and \lambda_{i,j} of order p^{k_{i,j}-h_{i,j}}, which is possible as each component is cyclic. Then \lambda has order |K/H|, and generate the quotient K/H.

Suppose that \hat{G}/\hat{H} is cyclic. Then it is generated by (x,b) \hat{H}. Apply \pi_2' and we get that bH= t^kH for some k, that is, b= t^kh with h\in H. Then let a=(\frac{x}{\psi(h)\alpha(h,t^k)})^{h^{-1}}; therefore, we have that [(a,t^k)] = [(\psi(h),h)][((\frac{x}{\psi(h)\alpha(h,t^k)})^{h^{-1}},t^k)] = [(x,ht^k)] = [(x,b)]

The only if is direct, and for the other implication we have that if [(\lambda,1)] \in \langle [(a,t)] \rangle then all elements of the form [(b,t^k)] can be written as [(b,t^k)] = [(b/(a^{1+t+\cdots + t^{k-1}} \prod_{h=1}^{k-1}(t,t^h),1 ] [(a,t)]^k \in \langle [(a,t)] \rangle

Now, [(\lambda,1)] \in \langle [(a,t)] \rangle if and only if \exists k such that (\lambda,1) = (a,t)^k (c,d) such that (c,d) \in \hat{H}. Therefore, we have that k=|G/H |, d=t^{-k}, and we have to solve \lambda \in Q a^{N_k(t)} \alpha_k^*(t)\alpha(t^k,t^{-k})\psi(t^{-k})^{t^k}. This is equivalent to solving the system

Note that \sigma : G \longrightarrow Gal(\mathbb{F}_q,\mathbb{F}_p) \simeq C_m, where C_m is the cyclic group of m elements. Therefore, \sigma (H) \leq C_m is cyclic and generated by some element \gamma \in Gal(\mathbb{F}_q,\mathbb{F}_p).

Let (a,b)\in N, (c,d) \in \hat{H}. Then

As (a,b) is in the normaliser, this should be in \hat{H}, which is equivalent to

Now, suppose that b=1. This is reduced to

Since this holds for every d \in H, it is equivalent to a a^{-\gamma_H} \in Q_{\hat{H}}. In a group setting, this condition simplifies to a^{\gamma_H - 1} \in Q_{\hat{H}}, from which we conclude that Q_N^{\gamma_H - 1} \subseteq Q_{\hat{H}}. To prove that it is maximal, let (x,1) \in \hat{G} such that x^{\gamma_H-1} \in Q_{\hat{H}}. Then (x,1)(c,d)(\frac{1}{x},1) = \left(\frac{x}{x^d}c,d \right). As x^{\gamma_H-1} \in Q_{\hat{H}}, in particular x^{d-1} \in Q_{\hat{H}} and \frac{x}{x^d}\in Q_{\hat{H}}, so \left(\frac{x}{x^d}c,d \right) = \left(\frac{x}{x^d},1) (c,d \right)\in H and (x,1) \in N.

For T, we can take a=\psi^*(b), and we get that

As c\in \psi(d) Q_N and Q_N^{1-b} \subseteq Q_{\hat{H}}, this can be reduced to

And therefore for all b\in T there exist such x, given by x=\psi^*(b). Finally, if for some b there exist x such that x^{1-d} \in Q_{H} \psi(d)^{1-b} \frac{\alpha(d,b)}{\alpha(b,d)}\quad \forall b \in H, then clearly (x,b) \in N and b\in T as we wanted to prove.

Let (c, t) \hat{H} \in \hat{K}/ \hat{H} be a generator, with tH a generator of K/H with t\in K of order |K/H| = [K:H]. Then, we have that

Therefore, if (a,b) \in E we have that

As a result t^{3^k} = t \mod H, so we have that [K:H] | 3^k-1. Consequently, we have that

Finally, Q_{\hat{H}} as well as in the previous proposition can be proved that it is determined by

and \pi_2(E) can be characterised as the set of elements b\in G such that the following system has a solution

for some k\in \{1,\cdots, [K:H] \} such that [K:H] | 3^k-1

It is a subgroup of \hat{G} as Z_\alpha^\sigma(G) is a subgroup of \hat{G} and Z_\alpha^\sigma(G) = \pi_2(A) To see that it is abelian, let (a,b),(c,d)\in A. Then

Now we will prove that it is maximal. Suppose there exists an abelian subgroup B\leq \hat{G} such that A\leq B. Then \mathbb{F}_q^* \times \{1\}\leq A \leq B, so B=\mathbb{F}_q^* \times \pi_2(B).

Take a\in \mathbb{F}_q^*, b \in \pi_2(B), so (1,b),(a,1) \in B, and as it has to be commutative

so a=a^b for all b\in \pi_2(B) and therefore B\leq \ker(\sigma). Now, for the 2-cocycle, take a,b\in \pi_2(B), then (1,a),(1,b) \in B, and as it has to be commutative:

so \alpha|_B is symmetric. But as A is maximal under this conditions, we have that B=A.

We will study \hat{H}'.

For all (a,b),(c,d) \in \hat{H}

If we take b=1 we get that \left(\frac{a}{a^d},1\right)\in \hat{H}' for all a\in Q_{\hat{H}}, d \in \pi_2(\hat{H}). In particular, we have \left(\frac{a}{\gamma(a)},1\right)\in \hat{H}'. For the second element, it is enough to take (\psi(b),b),( \psi(d),d) \in \hat{H}.

To prove that this generates the whole group, first note that \sigma(d) = \gamma^j for some j\in \mathbb{N}. In addition, \gamma(a)=a^{p^h} for some h\in \mathbb{N}. Therefore, we have that

So \left(\frac{a}{a^d},1 \right) \in \left \langle \left(\frac{c}{\gamma(c)}, 1 \right) ; c\in \mathbb{F}_q \right \rangle. Now, we have that

A direct application of proposition 19.

From 8, we have that \psi(a) must satisfy

however, as \hat{K}'\leq \hat{H}, we have that \frac{\psi(a)}{\psi(a)^b} ,\frac{\alpha(a,b)}{\alpha(b,a)} \in Q_{\hat{H}}, so the first condition is always true.