betaFeedback

WorksMathematics

Determinant characters and multiplicative solubility of finite skew braces

Abstract

Let BB be a finite skew brace. We prove that if the additive derived subgroup (B,+)′(B,+)' is nilpotent, then the multiplicative group (B,⋅)(B,\cdot) is soluble. In particular, this holds for finite skew braces with supersoluble or metabelian additive group.

Introduction

A skew brace is a triple (B,+,⋅)(B,+,\cdot) such that (B,+)(B,+) and (B,⋅)(B,\cdot) are groups and for every a,b,c∈Ba,b,c \in B:

a⋅(b+c)=(a⋅b)−a+(a⋅c)a \cdot(b+c)=(a\cdot b)-a+(a\cdot c)

The two groups are called the additive and multiplicative groups of BB, respectively. Introduced by Guarnieri and Vendramin [13] as a generalisation of Rump’s braces [17], skew braces arose in the study of set-theoretic solutions of the Yang–Baxter equation. They also connect this subject with regular subgroups of holomorphs and Hopf–Galois theory [18 Appendix A]. Throughout this paper, all groups and skew braces are finite. A central problem is to understand how the structure of either group constrains the other. This has been studied through solubility criteria [2, 11, 15, 20], restrictions involving simple groups [3, 19, 6], and comparisons of Sylow structure and derived subgroups [7, 8, 14]. When the additive group is abelian, the multiplicative group is soluble, as follows from the work of Etingof, Schedler and Soloviev [10]. Byott Byott [2, Theorem 1] proved the corresponding result for nilpotent additive groups in the language of Hopf–Galois theory; it was subsequently obtained for skew braces by Smoktunowicz and Vendramin [18 Corollary 2.23]. The natural extension to soluble additive groups was proposed by Byott [2] and formulated for skew braces in [18 Question 2.25] and [21 Problem 2.46]. Known as Byott’s conjecture, it remained open for more than a decade. Di Matteo, Ferrara and Trombetti [9] recently disproved it by constructing a skew brace with soluble additive group and insoluble multiplicative group. Our main result shows that nilpotency of the additive derived subgroup is sufficient.

Theorem A. Let BB be a finite skew brace. If (B,+)′(B,+)' is nilpotent, then (B,⋅)(B,\cdot) is soluble.

In particular, this extends the result for cyclic additive derived subgroups in [14]. The proof uses Byott’s description of a minimal counterexample [3 Theorem 3.2]. Its multiplicative group has a soluble normal subgroup RR with

(B,⋅)/R≃GL⁡3(2)r⋊T,(B,\cdot)/R \simeq\operatorname{GL}_3(2)^r \rtimes T,

where r≥1r \ge1 and T≤SrT \le S_r permutes the factors. We attach to this quotient a character ψ ⁣:(B,⋅)/R⟶F7×\psi\colon(B,\cdot)/R \longrightarrow\mathbb{F}_7^{\times}. For a suitable element z∈GL⁡3(2)rz \in\operatorname{GL}_3(2)^r, perfectness forces ψ(z)=1\psi(z)=1, whereas a determinant calculation gives ψ(z)=4\psi(z)=4. This contradiction excludes the minimal counterexample. Theorem A applies in particular to supersoluble groups, a class lying strictly between nilpotent and soluble groups. Their derived subgroups are nilpotent [16 Theorem 5.4.10]. It also applies to metabelian groups, whose derived subgroups are abelian.

Corollary. Let BB be a finite skew brace with supersoluble additive group. Then its multiplicative group is soluble.

Corollary. Let BB be a finite skew brace with metabelian additive group. Then its multiplicative group is soluble.

Theorem A applies to additive groups that are neither supersoluble nor metabelian. For instance, consider G=SL⁡2(3)≅Q8⋊C3G=\operatorname{SL}_2(3)\cong Q_8\rtimes C_3. Its derived subgroup is isomorphic to the quaternion group Q8Q_8, which is nilpotent but non-abelian. Moreover, GG is not supersoluble, since G/Z(G)≅A4G/Z(G)\cong A_4 is not supersoluble. Thus GG satisfies the hypothesis of Theorem A without satisfying either of the two conditions in the preceding corollaries.

The correspondence between skew braces and regular subgroups of holomorphs translates Theorem A into a restriction on Hopf–Galois types: a Galois extension with group GG admits a structure of type NN if and only if there is a skew brace with additive group NN and multiplicative group GG [12, 1, 18].

Corollary. Let L/KL/K be a finite Galois extension admitting a Hopf–Galois structure of type NN. If N′N' is nilpotent, then Gal⁡(L/K)\operatorname{Gal}(L/K) is soluble. In particular, this holds when NN is supersoluble or metabelian.

The paper is organised as follows. Section 2 collects the preliminaries, reduces Theorem A to the main obstruction, and establishes its principal consequences. Section 3 introduces determinant characters, while Section 4 uses normal additive Sylow subgroups to construct characters on multiplicative quotients. Finally, Section 5 proves the obstruction and completes the proof of Theorem A.

Preliminaries and reduction

Skew braces

2.1. Skew braces. A skew brace is a triple (B,+,⋅)(B,+,\cdot) such that both (B,+)(B,+) and (B,⋅)(B,\cdot) are groups and for all a,b,c∈Ba,b,c\in B the following compatibility condition holds:

a⋅(b+c)=(a⋅b)−a+(a⋅c)a \cdot(b+c)=(a\cdot b)-a+(a\cdot c)

The map λ:(B,⋅)→Aut⁡(B,+)\lambda:(B,\cdot)\to\operatorname{Aut}(B,+) sending a∈Ba\in B to λa\lambda_a where

λa(b)=−a+(a⋅b)\lambda_a(b)=-a+(a\cdot b)

is a well defined group homomorphism (see [13]). A subbrace of (B,+,⋅)(B,+,\cdot) is a subset C⊂BC\subset B such that (C,+)≤(B,+)(C,+)\leq(B,+) and (C,⋅)≤(B,⋅)(C,\cdot)\leq(B,\cdot). If (C,+)(C,+) is a characteristic subgroup of (B,+)(B,+), then CC is a subbrace of BB since for every x,y∈Cx,y\in C, x⋅y=x+λx(y)∈Cx\cdot y=x+\lambda_x(y)\in C. A subbrace LL of BB is called a left ideal if λb(L)⊆L\lambda_b(L)\subseteq L for every b∈Bb\in B; see [5]. A left ideal II of BB is called an ideal, and we write I⊴BI\trianglelefteq B, if (I,+)⊴(B,+)(I,+)\trianglelefteq(B,+), (I,⋅)⊴(B,⋅)(I,\cdot)\trianglelefteq(B,\cdot); see [18 Definition 5.1 and Lemma 5.5]. If II is an ideal, the additive and multiplicative cosets coincide, since

b⋅I=b+λb(I)=b+I(b∈B).b\cdot I=b+\lambda_b(I)=b+I\qquad(b\in B).

Thus the quotient set B/I={b+I:b∈B}B/I=\{b+I:b\in B\} inherits the operations

(a+I)+(b+I)=(a+b)+I,(a+I)+(b+I)=(a+b)+I,
(a+I)⋅(b+I)=(a⋅b)+I.(a+I)\cdot(b+I)=(a\cdot b)+I.

These operations are well defined and make B/IB/I a skew brace, called the quotient skew brace of BB by II. A map α:B⟶B\alpha:B\longrightarrow B is called an automorphism of the skew brace BB if α∈Aut⁡(B,+)∩Aut⁡(B,⋅)\alpha\in\operatorname{Aut}(B,+)\cap\operatorname{Aut}(B,\cdot). For a group GG, write G′G' for its derived subgroup, Φ(G)\Phi(G) for its Frattini subgroup, and NG(H)N_G(H) for the normalizer of a subgroup HH. For x,y∈Gx,y\in G we put Ad⁡x(y)=xyx−1\operatorname{Ad}_x(y)=xyx^{-1}.

Byott's reduction

As recalled in the introduction, Byott’s conjecture asks whether every finite skew brace with soluble additive group has soluble multiplicative group. Although the conjecture is false [9], Byott’s description of a minimal counterexample provides the starting point of our proof. We state his reduction in the language of skew braces.

Theorem 2.1 ([3 Theorem 3.2]). Let BB be a finite skew brace such that (B,+)(B,+) is soluble and (B,⋅)(B,\cdot) is insoluble. Suppose that every proper subbrace of BB has soluble multiplicative group. Then every maximal normal subgroup of (B,+)(B,+) has index 22, and there exist a soluble normal subgroup RR of (B,⋅)(B,\cdot), an integer r≥1r\geq1, and a soluble transitive subgroup T≤SrT\leq S_r such that

(B,⋅)/R≅GL⁡3(2)r⋊T,(B,\cdot)/R\cong\operatorname{GL}_3(2)^r\rtimes T,

where TT acts by permuting the rr factors.

Corollary 2.2. Under the hypotheses of Theorem 2.1, the group (B,+)/(B,+)′(B,+)/(B,+)' is a 22-group.

Proof. Let pp be a prime divisor of ∣(B,+)/(B,+)′∣\lvert(B,+)/(B,+)'\rvert. Since this quotient is abelian, it has a normal subgroup N/(B,+)′N/(B,+)' of index pp, where N⊴(B,+)N\trianglelefteq(B,+). Then [(B,+):N]=p[(B,+):N]=p, so NN is a maximal normal subgroup of (B,+)(B,+). Thus p=2p=2 by Theorem 2.1. □\square

Proof of Theorem A

We first reduce Theorem A to the following key obstruction, which will be proved in § 5.

Proposition 2.3. Let BB be a finite skew brace whose additive Sylow 33- and 77-subgroups are normal. Then (B,⋅)(B,\cdot) has no soluble normal subgroup RR such that (B,⋅)/R≃GL3(2)r⋊T(B,\cdot)/R \simeq\mathrm{GL}_{3}(2)^{r} \rtimes T, where r≥1r \ge1 and T≤SrT \le S_{r} permutes the factors.

Proof of Theorem A. Suppose otherwise, and choose a counterexample BB of minimal order. Since (B,+)′(B,+)' is nilpotent, (B,+)(B,+) is soluble. For every proper subbrace CC of BB, the group (C,+)′≤(B,+)′(C,+)' \le(B,+)' is nilpotent, so (C,⋅)(C,\cdot) is soluble by minimality. By Theorem 2.1, there exists a soluble normal subgroup RR of (B,⋅)(B,\cdot) with (B,⋅)/R≃GL3(2)r⋊T(B,\cdot)/R \simeq\mathrm{GL}_{3}(2)^{r} \rtimes T, and (B,+)/(B,+)′(B,+)/(B,+)' is a 22-group by Corollary 2.2. Thus, for p∈{3,7}p \in\{3,7\}, every Sylow pp-subgroup of (B,+)(B,+) lies in (B,+)′(B,+)' and is a Sylow pp-subgroup of (B,+)′(B,+)'. Since (B,+)′(B,+)' is nilpotent, this Sylow subgroup is unique and hence characteristic in (B,+)′(B,+)'. Moreover, (B,+)′(B,+)' is characteristic in (B,+)(B,+), so the Sylow subgroup is characteristic, and therefore normal, in (B,+)(B,+). Hence both additive Sylow subgroups are normal in (B,+)(B,+), contradicting Proposition 2.3. □\square

The proof of Theorem A thus reduces to Proposition 2.3. We construct a character ψ‾R,7,3 ⁣:(B,⋅)/R→F7×\overline{\psi}_{R,7,3} \colon(B,\cdot)/R \to\mathbb{F}_{7}^{\times} and evaluate it on a suitable element z∈GL3(2)rz \in\mathrm{GL}_{3}(2)^{r}. Since this subgroup is perfect, ψ‾R,7,3(z)=1\overline{\psi}_{R,7,3}(z)=1, whereas the determinant calculation gives ψ‾R,7,3(z)=4\overline{\psi}_{R,7,3}(z)=4. We can summarise the logical steps of the proof as follows:

(B,⋅)/R≃GL3(2)r⋊T⇝ψ‾R,7,3⇝1=4 in F7×(B,\cdot)/R \simeq\mathrm{GL}_{3}(2)^{r} \rtimes T \rightsquigarrow\overline{\psi}_{R,7,3} \rightsquigarrow1=4\text{ in }\mathbb{F}_{7}^{\times}

Consequences of Theorem A

Before proving Proposition 2.3, we record some immediate consequences of Theorem A.

Corollary 2.4. Let BB be a finite skew brace with supersoluble additive group. Then (B,⋅)(B,\cdot) is soluble.

Proof. The derived subgroup of a finite supersoluble group is nilpotent [16 Theorem 5.4.10]. Apply Theorem A. □\square

Corollary 2.5. Let BB be a finite skew brace with metabelian additive group. Then (B,⋅)(B,\cdot) is soluble.

Proof. The subgroup (B,+)′(B,+)' is abelian, hence nilpotent. Apply Theorem A. □\square

By the Greither–Pareigis correspondence [12] and Byott’s translation [1], a finite Galois extension L/KL/K admits a Hopf–Galois structure of type NN if and only if there is a skew brace with additive group isomorphic to NN and multiplicative group isomorphic to Gal⁡(L/K)\operatorname{Gal}(L/K); see [18 Appendix A].

Corollary 2.6. Let L/KL/K be a finite Galois extension admitting a Hopf–Galois structure of type NN. If N′N' is nilpotent, then Gal⁡(L/K)\operatorname{Gal}(L/K) is soluble.

Proof. The correspondence above gives a finite skew brace BB with (B,+)≅N(B,+)\cong N and (B,⋅)≅Gal⁡(L/K)(B,\cdot)\cong\operatorname{Gal}(L/K). Apply Theorem A. □\square

We now develop the determinant characters needed to prove Proposition 2.3.

Determinant characters

Determinants on finite pp-groups

Let PP be a finite non-trivial pp-group. By [16 Theorem 5.3.2] P/Φ(P)P/\Phi(P) is an elementary abelian pp-group and can be regarded as a vector space over Fp\mathbf F_p. Define the iterated Frattini subgroups by

P0=Φ0(P)=P,Pi+1=Φi+1(P)=Φ(Pi)(i≥0).P_0=\Phi^0(P)=P,\qquad P_{i+1}=\Phi^{i+1}(P)=\Phi(P_i)\quad(i\ge0).

Since the Frattini subgroup of a group is characteristic, each PiP_i is characteristic in PP. Moreover, the Frattini subgroup of a nontrivial finite pp-group is proper. Hence there exists an integer m≥0m\ge0 such that Pm=1P_m=1, and we obtain a characteristic series

P=P0≥P1≥⋯≥Pm=1.P=P_0\ge P_1\ge\cdots\ge P_m=1.

For every 0≤i<m0\le i<m, the quotient Vi=Pi/Pi+1=Pi/Φ(Pi)V_i=P_i/P_{i+1}=P_i/\Phi(P_i) is an elementary abelian pp-group, viewed as an Fp\mathbf F_p-vector space. Let α∈Aut⁡(P)\alpha\in\operatorname{Aut}(P). Since α\alpha preserves both PiP_i and Pi+1P_{i+1}, it induces an automorphism

αi:Vi⟶Vi,αi(xPi+1)=α(x)Pi+1.\alpha_i:V_i\longrightarrow V_i,\qquad\alpha_i(xP_{i+1})=\alpha(x)P_{i+1}.

We define

ΔP:Aut⁡(P)⟶Fp×,ΔP(α)=∏i=0m−1det⁡Fp(αi).(1)\Delta_P:\operatorname{Aut}(P)\longrightarrow\mathbf F_p^\times,\qquad\Delta_P(\alpha)=\prod_{i=0}^{m-1}\det_{\mathbf F_p}(\alpha_i). \tag*{(1)}

If P=1P=1, this is the empty product, with value 11. Equivalently, we may take the product over all i≥0i\ge0, assigning determinant 11 to every trivial factor. Finally, for α,β∈Aut⁡(P)\alpha,\beta\in\operatorname{Aut}(P), the automorphism induced by αβ\alpha\beta on ViV_i is αiβi\alpha_i\beta_i. The multiplicativity of the determinant therefore gives

ΔP(αβ)=∏i=0m−1det⁡Fp(αiβi)=(∏i=0m−1det⁡Fp(αi))(∏i=0m−1det⁡Fp(βi))=ΔP(α)ΔP(β).\Delta_P(\alpha\beta)=\prod_{i=0}^{m-1}\det_{\mathbf F_p}(\alpha_i\beta_i)=\left(\prod_{i=0}^{m-1}\det_{\mathbf F_p}(\alpha_i)\right)\left(\prod_{i=0}^{m-1}\det_{\mathbf F_p}(\beta_i)\right)=\Delta_P(\alpha)\Delta_P(\beta).

Hence ΔP\Delta_P is a group homomorphism.

We first establish some properties of the Δ\Delta-map that will be useful in the subsequent proofs.

Lemma 3.1. Let PP be a finite pp-group and α∈Aut⁡(P)\alpha\in\operatorname{Aut}(P). Then:

(1) For every isomorphism f:P→Qf:P\to Q, we have ΔQ(fαf−1)=ΔP(α)\Delta_Q(f\alpha f^{-1})=\Delta_P(\alpha).

(2) If I⊴PI\trianglelefteq P is α\alpha-invariant, then

ΔP(α)=ΔI(α∣I)ΔP/I(αˉ),(2)\Delta_P(\alpha)=\Delta_I(\alpha|_I)\Delta_{P/I}(\bar{\alpha}), \tag*{(2)}

where αˉ\bar{\alpha} is induced on P/IP/I.

Proof. For subgroups Y⊴X≤PY \trianglelefteq X \leq P satisfying α(X)=X\alpha(X)=X and α(Y)=Y\alpha(Y)=Y, denote by αX/Y\alpha_{X/Y} the induced automorphism

αX/Y:X/Y⟶X/Y,xY⟼α(x)Y.\alpha_{X/Y}:X/Y\longrightarrow X/Y,\qquad xY\longmapsto\alpha(x)Y.

For (1), since ff maps maximal subgroups of PP onto maximal subgroups of QQ, we have f(Φi(P))=Φi(Q)f(\Phi^i(P))=\Phi^i(Q) for every i≥0i\geq0. Thus ff induces the isomorphisms

fi:Φi(P)/Φi+1(P)⟶Φi(Q)/Φi+1(Q),xΦi+1(P)⟼f(x)Φi+1(Q).f_i:\Phi^i(P)/\Phi^{i+1}(P)\longrightarrow\Phi^i(Q)/\Phi^{i+1}(Q), \qquad x\Phi^{i+1}(P)\longmapsto f(x)\Phi^{i+1}(Q).

Here αi\alpha_i is an automorphism of Φi(P)/Φi+1(P)\Phi^i(P)/\Phi^{i+1}(P), whereas fiαifi−1f_i\alpha_i f_i^{-1} is an automorphism of Φi(Q)/Φi+1(Q)\Phi^i(Q)/\Phi^{i+1}(Q). For every y∈Φi(Q)y\in\Phi^i(Q),

(fiαifi−1)(yΦi+1(Q))=fi(α(f−1(y))Φi+1(P))=f(α(f−1(y)))Φi+1(Q)=(fαf−1)i(yΦi+1(Q)).\begin{aligned} (f_i\alpha_i f_i^{-1})(y\Phi^{i+1}(Q)) &=f_i\bigl(\alpha(f^{-1}(y))\Phi^{i+1}(P)\bigr)\\ &=f\bigl(\alpha(f^{-1}(y))\bigr)\Phi^{i+1}(Q)\\ &=(f\alpha f^{-1})_i(y\Phi^{i+1}(Q)). \end{aligned}

Hence (fαf−1)i=fiαifi−1(f\alpha f^{-1})_i=f_i\alpha_i f_i^{-1}, and therefore

ΔQ(fαf−1)=∏idet⁡(fiαifi−1)=∏idet⁡(αi)=ΔP(α).\Delta_Q(f\alpha f^{-1})=\prod_i\det(f_i\alpha_i f_i^{-1})=\prod_i\det(\alpha_i)=\Delta_P(\alpha).

For (2), we argue by induction on ∣P∣|P|. The cases I=1I=1 and I=PI=P are immediate. Suppose 1<I<P1<I<P and put F=Φ(P)F=\Phi(P). By [16 Theorem 5.3.2],

Φ(P/I)=(P/I)′(P/I)p=P′PpI/I=FI/I.\Phi(P/I)=(P/I)'(P/I)^p=P'P^pI/I=FI/I.

Consider the isomorphisms

θ:I/(I∩F)⟶FI/F,x(I∩F)⟼xF,\theta:I/(I\cap F)\longrightarrow FI/F,\qquad x(I\cap F)\longmapsto xF,
η:F/(I∩F)⟶FI/I,y(I∩F)⟼yI.\eta:F/(I\cap F)\longrightarrow FI/I,\qquad y(I\cap F)\longmapsto yI.

Their definitions give

θαI/(I∩F)θ−1=αFI/F,\theta\alpha_{I/(I\cap F)}\theta^{-1}=\alpha_{FI/F},
ηαF/(I∩F)η−1=αFI/I.\eta\alpha_{F/(I\cap F)}\eta^{-1}=\alpha_{FI/I}.

Since ∣I∣,∣F∣<∣P∣|I|,|F|<|P|, induction followed by (1) yields

ΔI(α)=ΔI∩F(α)ΔI/(I∩F)(α)=(1)ΔI∩F(α)ΔFI/F(α)=ΔI∩F(α)det⁡(αFI/F),ΔF(α)=ΔI∩F(α)ΔF/(I∩F)(α)=(1)ΔI∩F(α)ΔFI/I(α).\begin{aligned} \Delta_I(\alpha) &=\Delta_{I\cap F}(\alpha)\Delta_{I/(I\cap F)}(\alpha)\\ &\mathrel{\overset{(1)}{=}}\Delta_{I\cap F}(\alpha)\Delta_{FI/F}(\alpha)\\ &=\Delta_{I\cap F}(\alpha)\det(\alpha_{FI/F}),\\ \Delta_F(\alpha) &=\Delta_{I\cap F}(\alpha)\Delta_{F/(I\cap F)}(\alpha)\\ &\mathrel{\overset{(1)}{=}}\Delta_{I\cap F}(\alpha)\Delta_{FI/I}(\alpha). \end{aligned}

Here FI/FFI/F is elementary abelian. Since P/FP/F is an elementary abelian pp-group, we regard it as an Fp\mathbb{F}_p-vector space. The subgroup FI/FFI/F is an αP/F\alpha_{P/F}-invariant subspace. Choose a basis v1,…,vkv_1,\ldots,v_k of FI/FFI/F and extend it to a basis

B=(v1,…,vk,w1,…,wℓ)\mathcal{B}=(v_1,\ldots,v_k,w_1,\ldots,w_\ell)

of P/FP/F. Since αP/F(vj)∈FI/F\alpha_{P/F}(v_j)\in FI/F, the images of the first kk basis vectors have no components along the wiw_i. Hence

[αP/F]B=(AC0D)[\alpha_{P/F}]_{\mathcal{B}}= \begin{pmatrix} A & C\\ 0 & D \end{pmatrix}

where AA represents αFI/F\alpha_{FI/F}. The images of w1,…,wℓw_1,\ldots,w_\ell form a basis of (P/F)/(FI/F)≃P/FI(P/F)/(FI/F)\simeq P/FI.

On this quotient, the components along the viv_i vanish, so DD represents αP/FI\alpha_{P/FI}. Therefore

det⁡A=det⁡(αFI/F),det⁡D=det⁡(αP/FI).\det A=\det(\alpha_{FI/F}),\qquad\det D=\det(\alpha_{P/FI}).

Consequently,

det⁡(αP/F)=det⁡(AC0D)=det⁡Adet⁡D=det⁡(αFI/F)det⁡(αP/FI).\det(\alpha_{P/F})=\det\begin{pmatrix} A & C \\ 0 & D \end{pmatrix}=\det A\det D=\det(\alpha_{FI/F})\det(\alpha_{P/FI}).

Consequently,

ΔP(α)=det⁡(αP/F)ΔF(α)=det⁡(αFI/F)det⁡(αP/FI)ΔI∩F(α)ΔFI/I(α)=ΔI(α)det⁡(αP/FI)ΔFI/I(α)=ΔI(α)ΔP/I(α‾).\begin{aligned} \Delta_P(\alpha)&=\det(\alpha_{P/F})\Delta_F(\alpha)\\ &=\det(\alpha_{FI/F})\det(\alpha_{P/FI})\Delta_{I\cap F}(\alpha)\Delta_{FI/I}(\alpha)\\ &=\Delta_I(\alpha)\det(\alpha_{P/FI})\Delta_{FI/I}(\alpha)\\ &=\Delta_I(\alpha)\Delta_{P/I}(\overline{\alpha}). \end{aligned}

To justify the last equality, recall that the iterated Frattini series of P/IP/I begins with

P/I≥FI/I≥Φ(FI/I)≥⋯P/I\ge FI/I\ge\Phi(FI/I)\ge\cdots

with Φ(P/I)\Phi(P/I) under FI/IFI/I.

Let α‾0\overline{\alpha}_0 denote the automorphism induced on (P/I)/(FI/I)(P/I)/(FI/I). By definition, separating the first determinant from the remaining factors gives

ΔP/I(α‾)=det⁡(α‾0)ΔFI/I(α).\Delta_{P/I}(\overline{\alpha})=\det(\overline{\alpha}_0)\Delta_{FI/I}(\alpha).

Consider the isomorphism

ρ:(P/I)/(FI/I)⟶P/FI,(xI)(FI/I)⟼xFI.\rho:(P/I)/(FI/I)\longrightarrow P/FI,\qquad(xI)(FI/I)\longmapsto xFI.

For every x∈Px\in P, we have

ρ(α‾0((xI)(FI/I)))=ρ((α(x)I)(FI/I))=α(x)FI=αP/FI(ρ((xI)(FI/I))).\begin{aligned} \rho\bigl(\overline{\alpha}_0((xI)(FI/I))\bigr)&=\rho\bigl((\alpha(x)I)(FI/I)\bigr)\\ &=\alpha(x)FI\\ &=\alpha_{P/FI}\bigl(\rho((xI)(FI/I))\bigr). \end{aligned}

Hence

αP/FI=ρα‾0ρ−1,det⁡(αP/FI)=det⁡(α‾0).\alpha_{P/FI}=\rho\overline{\alpha}_0\rho^{-1},\qquad\det(\alpha_{P/FI})=\det(\overline{\alpha}_0).

Substituting into the preceding expression yields

ΔP/I(α‾)=det⁡(αP/FI)ΔFI/I(α).\Delta_{P/I}(\overline{\alpha})=\det(\alpha_{P/FI})\Delta_{FI/I}(\alpha).

Proposition 3.2. Let CC be a skew brace of pp-power order and let α\alpha be an automorphism of CC. Then

Δ(C,+)(α)=Δ(C,⋅)(α).(3)\Delta_{(C,+)}(\alpha)=\Delta_{(C,\cdot)}(\alpha). \tag*{(3)}

Proof. We argue by induction on ∣C∣|C|. Write C∗CC*C for the additive subgroup generated by x∗y=−x+x⋅y−yx*y=-x+x\cdot y-y, where x,y∈Cx,y\in C. By [5 Proposition 2.3], C∗CC*C is an ideal of CC and the quotient C/(C∗C)C/(C*C) is a trivial skew brace. Moreover, a skew brace of prime-power order is left nilpotent [5 Proposition 4.4]; thus C∗C≠CC*C\ne C if C≠1C\ne1. The automorphism α\alpha preserves C∗CC*C. Applying

Lemma 3.1(2) to the two group structures yields

Δ(C,+)(α)=Δ(C∗C,+)(α)Δ(C/(C∗C),+)(αˉ),\Delta_{(C,+)}(\alpha)=\Delta_{(C*C,+)}(\alpha)\Delta_{(C/(C*C),+)}(\bar{\alpha}),
Δ(C,⋅)(α)=Δ(C∗C,⋅)(α)Δ(C/(C∗C),⋅)(αˉ).\Delta_{(C,\cdot)}(\alpha)=\Delta_{(C*C,\cdot)}(\alpha)\Delta_{(C/(C*C),\cdot)}(\bar{\alpha}).

The first factors agree by induction and the second agree because the quotient is trivial. □\square

A character associated with a soluble normal subgroup

Lemma 3.3. Let RR be a soluble normal subgroup of a finite group GG, and let pp be a prime. There exists a homomorphism βR,p:G→Fp×\beta_{R,p}:G\to\mathbb{F}_p^{\times} such that, for every P∈Syl⁡p(G)P\in\operatorname{Syl}_p(G) and h∈NG(P)h\in N_G(P),

βR,p(h)=ΔP∩R(Ad⁡h∣P∩R).(4)\beta_{R,p}(h)=\Delta_{P\cap R}(\operatorname{Ad}_h\mid_{P\cap R}). \tag*{(4)}

Moreover,

ΔP(Ad⁡h∣P)=βR,p(h)ΔPR/R(Ad⁡hR∣PR/R).(5)\Delta_P(\operatorname{Ad}_h\mid_P)=\beta_{R,p}(h)\Delta_{PR/R}(\operatorname{Ad}_{hR}\mid_{PR/R}). \tag*{(5)}

Proof. Refining the derived series of RR by the primary and power filtrations of its abelian factors, choose a series 1=Rs≤⋯≤R0=R1=R_s\leq\cdots\leq R_0=R, where Ri⊴GR_i\unlhd G and Ri/Ri+1≅FqidiR_i/R_{i+1}\cong\mathbb{F}_{q_i}^{d_i}, with each qiq_i a prime. For g∈Gg\in G, define

ρi(g):Ri/Ri+1⟶Ri/Ri+1,xRi+1⟼gxg−1Ri+1,\rho_i(g):R_i/R_{i+1}\longrightarrow R_i/R_{i+1},\qquad xR_{i+1}\longmapsto gxg^{-1}R_{i+1},

and put

βR,p(g)=∏0≤i≤sqi=pdet⁡Fp(ρi(g)).\beta_{R,p}(g)=\prod_{\substack{0\leq i\leq s\\q_i=p}}\det_{\mathbb{F}_p}(\rho_i(g)).

Since ρi(gg′)=ρi(g)ρi(g′)\rho_i(gg')=\rho_i(g)\rho_i(g'), we have βR,p(gg′)=βR,p(g)βR,p(g′)\beta_{R,p}(gg')=\beta_{R,p}(g)\beta_{R,p}(g'). Fix P∈Syl⁡p(G)P\in\operatorname{Syl}_p(G) and h∈NG(P)h\in N_G(P). Since Ri⊴GR_i\unlhd G, we have P∩Ri∈Syl⁡p(Ri)P\cap R_i\in\operatorname{Syl}_p(R_i). Therefore

∣P∩RiP∩Ri+1∣=∣Ri∣p∣Ri+1∣p={pdi,qi=p,1,qi≠p.\left|\frac{P\cap R_i}{P\cap R_{i+1}}\right|=\frac{|R_i|_p}{|R_{i+1}|_p}= \begin{cases} p^{d_i}, & q_i=p,\\ 1, & q_i\ne p. \end{cases}

Let σi\sigma_i be the automorphism induced by Ad⁡h\operatorname{Ad}_h on (P∩Ri)/(P∩Ri+1)(P\cap R_i)/(P\cap R_{i+1}), namely

σi(x(P∩Ri+1))=hxh−1(P∩Ri+1).\sigma_i\bigl(x(P\cap R_{i+1})\bigr)=hxh^{-1}(P\cap R_{i+1}).

If qi≠pq_i\ne p, this factor is trivial and det⁡(σi)=1\det(\sigma_i)=1. If qi=pq_i=p, the map

θi:P∩RiP∩Ri+1⟶Ri/Ri+1,x(P∩Ri+1)⟼xRi+1,\theta_i:\frac{P\cap R_i}{P\cap R_{i+1}}\longrightarrow R_i/R_{i+1},\qquad x(P\cap R_{i+1})\longmapsto xR_{i+1},

is injective, and both groups have order pdip^{d_i}; hence it is an isomorphism. Moreover,

θiσi(x(P∩Ri+1))=hxh−1Ri+1\theta_i\sigma_i\bigl(x(P\cap R_{i+1})\bigr)=hxh^{-1}R_{i+1}
θiσi(x(P∩Ri+1))=ρi(h)θi(x(P∩Ri+1)).\phantom{\theta_i\sigma_i\bigl(x(P\cap R_{i+1})\bigr)} =\rho_i(h)\theta_i\bigl(x(P\cap R_{i+1})\bigr).

Thus ρi(h)=θiσiθi−1\rho_i(h)=\theta_i\sigma_i\theta_i^{-1} and det⁡(σi)=det⁡(ρi(h))\det(\sigma_i)=\det(\rho_i(h)). Iterating Lemma 3.1(2) along 1=P∩Rs≤⋯≤P∩R0=P∩R1=P\cap R_s\leq\cdots\leq P\cap R_0=P\cap R now gives

ΔP∩R(Ad⁡h∣P∩R)=∏i=0s−1det⁡(σi)=∏0≤i<sqi=pdet⁡(ρi(h))=βR,p(h),(6)\Delta_{P\cap R}(\operatorname{Ad}_h|_{P\cap R})=\prod_{i=0}^{s-1}\det(\sigma_i)=\prod_{\substack{0\leq i<s\\q_i=p}}\det(\rho_i(h))=\beta_{R,p}(h), \tag*{(6)}

proving (3.4). Finally, the isomorphism

θ:P/(P∩R)⟶PR/R,x(P∩R)⟼xR,\theta:P/(P\cap R)\longrightarrow PR/R,\qquad x(P\cap R)\longmapsto xR,

satisfies θAd⁡‾hθ−1=Ad⁡hR∣PR/R\theta\overline{\operatorname{Ad}}_h\theta^{-1}=\operatorname{Ad}_{hR}|_{PR/R}, where Ad⁡‾h\overline{\operatorname{Ad}}_h is induced on P/(P∩R)P/(P\cap R). Applying Lemma 3.1(2), then (1) and (3.4), yields

ΔP(Ad⁡h∣P)=ΔP∩R(Ad⁡h∣P∩R)ΔP/(P∩R)(Ad⁡‾h)=βR,p(h)ΔPR/R(Ad⁡hR∣PR/R).\begin{aligned} \Delta_P(\operatorname{Ad}_h|_P)=\Delta_{P\cap R}(\operatorname{Ad}_h|_{P\cap R})\Delta_{P/(P\cap R)}(\overline{\operatorname{Ad}}_h) \\ &=\beta_{R,p}(h)\Delta_{PR/R}(\operatorname{Ad}_{hR}|_{PR/R}). \end{aligned}

Normal additive Sylow subgroups and quotient characters

Normalizers of Sylow subbraces

Normalizers of Sylow subbraces. We study multiplicative normalizers when two normal additive Sylow subgroups have coprime orders and hence commute.

Lemma 4.1. Let BB be a finite skew brace and let p,qp,q be distinct primes dividing ∣B∣|B|. Suppose that (P,+)(P,+) and (Q,+)(Q,+) are normal Sylow pp- and qq-subgroups of (B,+)(B,+), respectively. Then:

  1. N(B,⋅)((P,⋅))N_{(B,\cdot)}((P,\cdot)) is a subbrace of BB.

  2. N(B,⋅)((P,⋅))∩(Q,⋅)∈Syl⁡q(N(B,⋅)((P,⋅)))N_{(B,\cdot)}((P,\cdot))\cap(Q,\cdot)\in\operatorname{Syl}_q(N_{(B,\cdot)}((P,\cdot))).

  3. For every t∈N(B,⋅)((P,⋅))∩(Q,⋅)t\in N_{(B,\cdot)}((P,\cdot))\cap(Q,\cdot), the restriction λt∣P\lambda_t|_P is an automorphism of the skew brace PP, and

λt(x)=t⋅x⋅t−1(x∈P).(7)\lambda_t(x)=t\cdot x\cdot t^{-1}\qquad(x\in P). \tag*{(7)}

where t−1t^{-1} denotes the inverse in (B,⋅)(B,\cdot).

Proof. (1) Since (P,+)(P,+) and (Q,+)(Q,+) are normal Sylow subgroups of (B,+)(B,+), they are characteristic in (B,+)(B,+). Thus PP and QQ are left ideals and subbraces of BB. By [4 Theorem A(2)], the intersection N(B,+)((P,+))∩N(B,⋅)((P,⋅))N_{(B,+)}((P,+))\cap N_{(B,\cdot)}((P,\cdot)) is a subbrace of BB. Since (P,+)⊴(B,+)(P,+)\trianglelefteq(B,+), its additive normalizer is (B,+)(B,+). Hence this intersection is precisely N(B,⋅)((P,⋅))N_{(B,\cdot)}((P,\cdot)).

(2) Write ∣N(B,⋅)((P,⋅))∣=qam|N_{(B,\cdot)}((P,\cdot))|=q^a m, where q∤mq\nmid m. By (1), the additive quotient map restricts to a homomorphism

π:(N(B,⋅)((P,⋅)),+)⟶(B,+)/(Q,+),b⟼b+Q.\pi:\left(N_{(B,\cdot)}((P,\cdot)),+\right)\longrightarrow(B,+)/(Q,+),\qquad b\longmapsto b+Q.

Its kernel is ker⁡π=N(B,⋅)((P,⋅))∩Q\ker\pi=N_{(B,\cdot)}((P,\cdot))\cap Q. Since the normalizer and QQ are subbraces, this intersection is also a multiplicative subgroup. This is a qq-group, whereas im⁡π\operatorname{im}\pi has order prime to qq. The first isomorphism theorem therefore gives ∣ker⁡π∣=qa\lvert\ker\pi\rvert=q^{a} proving (2).

(3) Take t∈N(B,⋅)((P,⋅)∩(Q,⋅))t\in N_{(B,\cdot)}((P,\cdot)\cap(Q,\cdot)) and x∈Px\in P and put y=t⋅x⋅t−1y=t\cdot x\cdot t^{-1}. Then y⋅t=t⋅xy\cdot t=t\cdot x and so y+λy(t)=t+λt(x)y+\lambda_y(t)=t+\lambda_t(x). Note that y,λt(x)∈Py,\lambda_t(x)\in P and t,λy(t)∈Qt,\lambda_y(t)\in Q. Since (P,+)(P,+) and (Q,+)(Q,+) centralize each other, we may rewrite the equality as y+λy(t)=λt(x)+ty+\lambda_y(t)=\lambda_t(x)+t. Comparing the unique components in PP and QQ gives y=λt(x)y=\lambda_t(x), proving (4.1). Thus λt∣P\lambda_t|_P is an additive automorphism and agrees with a multiplicative automorphism of (P,⋅)(P,\cdot). It is therefore an automorphism of the skew brace PP. □\square

Constructing the quotient character

We first record the group-theoretic lemma used to make the character descend to the multiplicative quotient.

Lemma 4.2. Let GG be a finite group, let qq be a prime, and let AA be an abelian qq-group. Let ψ:G→A\psi:G\to A be a homomorphism and R⊴GR\trianglelefteq G. Suppose p≠qp\ne q, P∈Syl⁡p(G)P\in\operatorname{Syl}_p(G) and Q∈Syl⁡q(NG(P))Q\in\operatorname{Syl}_q(N_G(P)). If ψ\psi is trivial on R∩QR\cap Q, then it is trivial on RR.

Proof. Since ψ(P)=1\psi(P)=1, the subgroup PP is a Sylow pp-subgroup of RP∩ker⁡ψRP\cap\ker\psi. The latter is normal in RPRP, so the Frattini argument (see [16 Theorem 5.2.14]) yields RP=(RP∩ker⁡ψ)NRP(P)RP=(RP\cap\ker\psi)N_{RP}(P). Since NRP(P)=NR(P)PN_{RP}(P)=N_R(P)P, we obtain

ψ(R)=ψ(RP)=ψ(NRP(P))=ψ(NR(P)).\psi(R)=\psi(RP)=\psi(N_{RP}(P))=\psi(N_R(P)).

If this image were nontrivial, choose u∈NR(P)u\in N_R(P) with ψ(u)≠1\psi(u)\ne1. Write ∣u∣=qam\lvert u\rvert=q^{a}m, where (q,m)=1(q,m)=1, and set v=umv=u^m. Then vv is a qq-element of NR(P)N_R(P) and ψ(v)=ψ(u)m≠1\psi(v)=\psi(u)^m\ne1. By Sylow’s theorem, some conjugate tt of vv by an element of NG(P)N_G(P) belongs to QQ. The normality of RR gives t∈R∩Qt\in R\cap Q, whereas ψ(t)=ψ(v)≠1\psi(t)=\psi(v)\ne1, a contradiction. □\square

Definition 4.3. Let BB be a finite skew brace, let pp be a prime dividing ∣B∣\lvert B\rvert, and suppose that (P,+)(P,+) is a normal Sylow pp-subgroup of (B,+)(B,+). Let RR be a soluble normal subgroup of (B,⋅)(B,\cdot). Define

χR,p:(B,⋅)⟶Fp×,χR,p(b)=Δ(P,+)(λb∣P) βR,p(b)−1.\chi_{R,p}:(B,\cdot)\longrightarrow\mathbb{F}_p^\times,\qquad\chi_{R,p}(b)=\Delta_{(P,+)}(\lambda_b|_P)\,\beta_{R,p}(b)^{-1}.

Since (P,+)(P,+) is characteristic in (B,+)(B,+), the restriction λb∣P\lambda_b|_P belongs to Aut⁡(P,+)\operatorname{Aut}(P,+) for every b∈Bb\in B. Thus χR,p\chi_{R,p} is well defined. Both factors in its definition are homomorphisms, so χR,p\chi_{R,p} is a homomorphism.

Lemma 4.4. Let BB be a finite skew brace, and let p,qp,q be distinct prime divisors of ∣B∣\lvert B\rvert such that q∣p−1q\mid p-1. Suppose that (P,+)(P,+) and (Q,+)(Q,+) are normal Sylow pp- and qq-subgroups of (B,+)(B,+), respectively. Let RR be a soluble normal subgroup of (B,⋅)(B,\cdot). Then, for every t∈N(B,⋅)((P,⋅))∩(Q,⋅)t\in N_{(B,\cdot)}((P,\cdot))\cap(Q,\cdot),

χR,p(t)=Δ(P,⋅)R/R(Ad⁡tR).(8)\chi_{R,p}(t)=\Delta_{(P,\cdot)R/R}(\operatorname{Ad}_tR). \tag*{(8)}

Write p−1=qamp-1=q^{a}m, where a≥1a\ge1 and gcd⁡(q,m)=1\gcd(q,m)=1. The character ψR,p,q=χR,pm\psi_{R,p,q}=\chi_{R,p}^{m} is trivial on RR and induces

ψ‾R,p,q:(B,⋅)/R⟶Fp×,ψ‾R,p,q(bR)=χR,p(b)m.\overline{\psi}_{R,p,q}:(B,\cdot)/R\longrightarrow\mathbb{F}_p^\times,\qquad\overline{\psi}_{R,p,q}(bR)=\chi_{R,p}(b)^m.

Proof. Take t∈N(B,⋅)((P,⋅))∩(Q,⋅)t \in N_{(B,\cdot)}((P,\cdot)) \cap(Q,\cdot). By Lemma 4.1, the restriction λt∣P\lambda_t|_P is an automorphism of the skew brace PP and agrees with multiplicative conjugation by tt. Proposition 3.2 and Lemma 3.3 give

Δ(P,+)(λt∣P)=Δ(P,⋅)(Ad⁡t∣P)=βR,p(t) Δ(P,⋅)R/R(Ad⁡tR).\Delta_{(P,+)}(\lambda_t|_P)=\Delta_{(P,\cdot)}(\operatorname{Ad}_t|_P) =\beta_{R,p}(t)\,\Delta_{(P,\cdot)_R/R}(\operatorname{Ad}_{tR}).

Dividing by βR,p(t)\beta_{R,p}(t) proves (4.2). Since ψR,p,q(b)qa=χR,p(b)p−1=1\psi_{R,p,q}(b)^{q^a}=\chi_{R,p}(b)^{p-1}=1 for every b∈Bb \in B, the image of ψR,p,q\psi_{R,p,q} is an abelian qq-group. If, in addition, t∈Rt \in R, then tR=RtR=R, so (4.2) gives χR,p(t)=1\chi_{R,p}(t)=1, and hence ψR,p,q(t)=1\psi_{R,p,q}(t)=1. Thus ψR,p,q\psi_{R,p,q} is trivial on R∩N(B,⋅)((P,⋅))∩(Q,⋅)R \cap N_{(B,\cdot)}((P,\cdot)) \cap(Q,\cdot). By Lemma 4.1(2), the group N(B,⋅)((P,⋅))∩(Q,⋅)N_{(B,\cdot)}((P,\cdot)) \cap(Q,\cdot) is a Sylow qq-subgroup of N(B,⋅)((P,⋅))N_{(B,\cdot)}((P,\cdot)). Lemma 4.2 therefore gives R≤ker⁡ψR,p,qR \leq\ker\psi_{R,p,q}, yielding the stated quotient character. □\square

Summarizing the preceding results, we obtain the following construction of determinant characters and their factorization through the multiplicative quotient.

R⊴(B,⋅) soluble,(P,+)⊴(B,+)↓χR,p:(B,⋅)⟶Fp×↓q∣p−1,(Q,+)⊴(B,+)↓[P,Q]+={0}ψR,p,q=χR,pm:(B,⋅)⟶Fp×↓R≤ker⁡ψR,p,qψ‾R,p,q:(B,⋅)/R⟶Fp×\begin{gathered} R \unlhd(B,\cdot)\ \text{soluble},\quad(P,+)\unlhd(B,+)\\ \downarrow\\ \chi_{R,p}:(B,\cdot)\longrightarrow\mathbb{F}_p^\times\\ \downarrow\quad q\mid p-1,\quad(Q,+)\unlhd(B,+)\\ \phantom{\downarrow}\quad[P,Q]_+=\{0\}\\ \psi_{R,p,q}=\chi_{R,p}^{m}:(B,\cdot)\longrightarrow\mathbb{F}_p^\times\\ \downarrow\quad R\leq\ker\psi_{R,p,q}\\ \overline{\psi}_{R,p,q}:(B,\cdot)/R\longrightarrow\mathbb{F}_p^\times \end{gathered}

We first recall a group-theoretic fact about normalizers in quotients, which will be used to evaluate the quotient character on the normalizer of a multiplicative Sylow subgroup.

Lemma 4.5. Let R⊴GR \unlhd G and P∈Syl⁡p(G)P \in\operatorname{Syl}_p(G). The quotient map induces a surjective homomorphism

φ:NG(P)⟶NG/R(PR/R),g⟼gR.\varphi:N_G(P)\longrightarrow N_{G/R}(PR/R),\qquad g\longmapsto gR.

Proof. Let gR∈NG/R(PR/R)gR \in N_{G/R}(PR/R). Then g(PR)g−1=PRg(P R)g^{-1}=P R, so gPg−1gPg^{-1} and PP are Sylow pp-subgroups of PRP R. Hence some element ur∈PRur \in P R, with u∈Pu \in P and r∈Rr \in R, conjugates gPg−1gPg^{-1} to PP. Since uu normalizes PP, also rr conjugates gPg−1gPg^{-1} to PP. Consequently rg∈NG(P)rg \in N_G(P) and (rg)R=gR(rg)R=gR, proving surjectivity. □\square

Lemma 4.6. Assume the hypotheses and notation of Lemma 4.4. For every z∈N(B,⋅)/R((P,⋅)R/R)z \in N_{(B,\cdot)/R}((P,\cdot)R/R), we have

ψ‾R,p,q(z)=Δ(P,⋅)R/R(Ad⁡z)m.(9)\overline{\psi}_{R,p,q}(z)=\Delta_{(P,\cdot)R/R}(\operatorname{Ad}_z)^m. \tag*{(9)}

Proof. Recall that p−1=qamp-1=q^{a}m, where gcd⁡(q,m)=1\operatorname{gcd}(q,m)=1. Define

Φ ⁣:N(B,⋅)/R((P,⋅)R/R)⟶{u∈Fp×:uqa=1},z⟼ψ‾R,p,q(z) Δ(P,⋅)R/R(Ad⁡z)−m.\Phi\colon N_{(B,\cdot)}/R((P,\cdot)R/R)\longrightarrow\{u\in\mathbb{F}_{p}^{\times}:u^{q^{a}}=1\}, \qquad z\longmapsto\overline{\psi}_{R,p,q}(z)\,\Delta_{(P,\cdot)R/R}(\operatorname{Ad}_{z})^{-m}.

Both factors are homomorphisms into the indicated abelian qq-group, so Φ\Phi is a homomorphism. By Lemma 4.5, the quotient map induces a surjection

N(B,⋅)((P,⋅))⟶N(B,⋅)/R((P,⋅)R/R),t⟼tR.N_{(B,\cdot)}((P,\cdot))\longrightarrow N_{(B,\cdot)}/R((P,\cdot)R/R), \qquad t\longmapsto tR.

By Lemma 4.1(2), the subgroup N(B,⋅)((P,⋅))∩(Q,⋅)N_{(B,\cdot)}((P,\cdot))\cap(Q,\cdot) is a Sylow qq-subgroup of N(B,⋅)((P,⋅))N_{(B,\cdot)}((P,\cdot)). Its image is therefore a Sylow qq-subgroup of the domain of Φ\Phi. For every t∈N(B,⋅)((P,⋅))∩(Q,⋅)t\in N_{(B,\cdot)}((P,\cdot))\cap(Q,\cdot), Lemma 4.4 gives

Φ(tR)=χR,p(t)m Δ(P,⋅)R/R(Ad⁡tR)−m=1.\Phi(tR)=\chi_{R,p}(t)^{m}\,\Delta_{(P,\cdot)R/R}(\operatorname{Ad}_{tR})^{-m}=1.

Thus ker⁡Φ\ker\Phi contains a Sylow qq-subgroup of its domain, so its index is prime to qq. This index equals ∣im⁡Φ∣|\operatorname{im}\Phi|, which is a power of qq. Hence Φ\Phi is trivial, proving (9).

Proof of the main obstruction

The following lemma provides the element and the determinant calculation needed to complete the proof of Proposition 2.3.

Lemma 5.1. Let r≥1r\geq1, T≤SrT\leq S_{r}, and let W=GL⁡3(2)r⋊TW=\operatorname{GL}_{3}(2)^{r}\rtimes T, with TT permuting the factors. For every P∈Syl⁡7(W)P\in\operatorname{Syl}_{7}(W), there exists an element z∈GL⁡3(2)r∩NW(P)z\in\operatorname{GL}_{3}(2)^{r}\cap N_{W}(P) of order 33 such that ΔP(Ad⁡z)=2\Delta_{P}(\operatorname{Ad}_{z})=2 in F7×\mathbb{F}_{7}^{\times}.

Proof. We first prove the lemma for a particular Sylow 77-subgroup. Regard F8\mathbb{F}_{8} as a 33-dimensional vector space over F2\mathbb{F}_{2}. Choose a generator ξ\xi of F8×\mathbb{F}_{8}^{\times} and define u ⁣:F8⟶F8u\colon\mathbb{F}_{8}\longrightarrow\mathbb{F}_{8} by u(x)=ξxu(x)=\xi x for every x∈F8x\in\mathbb{F}_{8}. Then u∈GL⁡3(2)u\in\operatorname{GL}_{3}(2) and ∣u∣=7|u|=7. Since ∣GL⁡3(2)∣=168|\operatorname{GL}_{3}(2)|=168, the subgroup U=⟨u⟩U=\langle u\rangle is a Sylow 77-subgroup of GL⁡3(2)\operatorname{GL}_{3}(2). Choose V∈Syl⁡7(T)V\in\operatorname{Syl}_{7}(T) and put P0=Ur⋊VP_{0}=U^{r}\rtimes V. This subgroup is well defined because TT permutes the coordinates of UrU^{r}. Moreover,

∣P0∣=7r∣V∣=7r∣T∣7=∣W∣7,|P_{0}|=7^{r}|V|=7^{r}|T|_{7}=|W|_{7},

so P0∈Syl⁡7(W)P_{0}\in\operatorname{Syl}_{7}(W). Consider the F2\mathbb{F}_{2}-linear Frobenius map s(x)=x2s(x)=x^{2}. Since s3(x)=x8=xs^{3}(x)=x^{8}=x and s≠1s\neq1, it has order 33, with s−1(x)=x4s^{-1}(x)=x^{4}. For every j∈Zj\in\mathbb{Z} and x∈F8x\in\mathbb{F}_{8} we have

(sujs−1)(x)=s(ξjx4)=ξ2jx8=ξ2jx=u2j(x).(su^{j}s^{-1})(x)=s(\xi^{j}x^{4})=\xi^{2j}x^{8}=\xi^{2j}x=u^{2j}(x).

Thus ss normalizes U=⟨u⟩U=\langle u\rangle and Ad⁡s(uj)=u2j\operatorname{Ad}_{s}(u^{j})=u^{2j}. Since UU is cyclic of order 77, we have Φ(U)=1\Phi(U)=1. Hence the only nontrivial Frattini factor is U/1U/1. Consider the isomorphism

θ ⁣:U⟶(F7,+),θ(uj)=[j]7,\theta\colon U\longrightarrow(\mathbb{F}_{7},+),\qquad\theta(u^{j})=[j]_{7},

and the automorphism

μ2 ⁣:(F7,+)⟶(F7,+),μ2(a)=2a.\mu_{2}\colon(\mathbb{F}_{7},+)\longrightarrow(\mathbb{F}_{7},+),\qquad\mu_{2}(a)=2a.

For every uj∈Uu^j \in U, we have

θ(Ad⁡s(uj))=θ(u2j)=[2j]7=μ2([j]7)=μ2(θ(uj)).\theta(\operatorname{Ad}_{s}(u^j))=\theta(u^{2j})=[2j]_7=\mu_2([j]_7)=\mu_2(\theta(u^j)).

Hence θAd⁡s=μ2θ\theta\operatorname{Ad}_s=\mu_2\theta, or equivalently, θAd⁡sθ−1=μ2\theta\operatorname{Ad}_s\theta^{-1}=\mu_2. Lemma 3.1(1), applied to the isomorphism θ\theta, therefore gives

ΔU(Ad⁡s)=Δ(F7,+)(θAd⁡sθ−1)=Δ(F7,+)(μ2).\Delta_U(\operatorname{Ad}_s)=\Delta_{(\mathbb{F}_7,+)}(\theta\operatorname{Ad}_s\theta^{-1})=\Delta_{(\mathbb{F}_7,+)}(\mu_2).

Since Φ((F7,+))={0}\Phi((\mathbb{F}_7,+))=\{0\}, the definition of Δ\Delta now yields

Δ(F7,+)(μ2)=det⁡F7(μ2)=det⁡F7(2)=2.\Delta_{(\mathbb{F}_7,+)}(\mu_2)=\det_{\mathbb{F}_7}(\mu_2)=\det_{\mathbb{F}_7}(2)=2.

Thus ΔU(Ad⁡s)=2\Delta_U(\operatorname{Ad}_s)=2. Consider the natural permutation action of V≤T≤SrV\leq T\leq S_r on {1,…,r}\{1,\ldots,r\}, given by v⋅i=v(i)v\cdot i=v(i). Choose an orbit O\mathcal{O} of this action. By the orbit–stabilizer theorem ∣O∣=[V:Stab⁡V(i)]=7k|\mathcal{O}|=[V:\operatorname{Stab}_V(i)]=7^k for any i∈Oi\in\mathcal{O} and some k≥0k\geq0, since VV is a 7-group. Define

z0=(z1,…,zr),zi={s,i∈O,1,i∉O.z_0=(z_1,\ldots,z_r),\qquad z_i=\begin{cases}s,&i\in\mathcal{O},\\1,&i\notin\mathcal{O}.\end{cases}

Then z0z_0 has order 3 and normalizes UrU^r. For every v∈Vv\in V we have

vz0v−1=(zv−1(1),…,zv−1(r))=(z1,…,zr)=z0,vz_0v^{-1}=(z_{v^{-1}(1)},\ldots,z_{v^{-1}(r)})=(z_1,\ldots,z_r)=z_0,

because ziz_i is constant on each VV-orbit. Thus z0z_0 centralizes VV, and consequently z0∈GL⁡3(2)r∩NW(P0)z_0\in\operatorname{GL}_3(2)^r\cap N_W(P_0). For every (ua1,…,uar)∈Ur(u^{a_1},\ldots,u^{a_r})\in U^r, conjugation is computed coordinatewise:

Ad⁡z0(ua1,…,uar)=(z1ua1z1−1,…,zruarzr−1)=(ud1a1,…,udrar),\operatorname{Ad}_{z_0}(u^{a_1},\ldots,u^{a_r})=(z_1u^{a_1}z_1^{-1},\ldots,z_ru^{a_r}z_r^{-1}) =(u^{d_1a_1},\ldots,u^{d_ra_r}),

where

di={2,i∈O,1,i∉O.d_i=\begin{cases}2,&i\in\mathcal{O},\\1,&i\notin\mathcal{O}.\end{cases}

Here we have used suas−1=u2asu^as^{-1}=u^{2a}. Consider the group isomorphism

θr:Ur⟶(F7r,+),(ua1,…,uar)⟼([a1]7,…,[ar]7),\theta_r:U^r\longrightarrow(\mathbb{F}_7^r,+),\qquad(u^{a_1},\ldots,u^{a_r})\longmapsto([a_1]_7,\ldots,[a_r]_7),

and the linear automorphism

μ:F7r⟶F7r,μ(a1,…,ar)=(d1a1,…,drar).\mu:\mathbb{F}_7^r\longrightarrow\mathbb{F}_7^r,\qquad\mu(a_1,\ldots,a_r)=(d_1a_1,\ldots,d_ra_r).

For every (ua1,…,uar)∈Ur(u^{a_1},\ldots,u^{a_r})\in U^r, we have

θr(Ad⁡z0(ua1,…,uar))=θr(ud1a1,…,udrar)=([d1a1]7,…,[drar]7)=μ([a1]7,…,[ar]7)=μ(θr(ua1,…,uar)).\begin{aligned} \theta_r(\operatorname{Ad}_{z_0}(u^{a_1},\ldots,u^{a_r}))&=\theta_r(u^{d_1a_1},\ldots,u^{d_ra_r})\\ &=([d_1a_1]_7,\ldots,[d_ra_r]_7)\\ &=\mu([a_1]_7,\ldots,[a_r]_7)\\ &=\mu(\theta_r(u^{a_1},\ldots,u^{a_r})). \end{aligned}

Thus θr(Ad⁡z0∣Ur)θr−1=μ\theta_r(\operatorname{Ad}_{z_0}|_{U^r})\theta_r^{-1}=\mu. By Lemma 3.1(1),

ΔUr(Ad⁡z0∣Ur)=Δ(F7r,+)(θr(Ad⁡z0∣Ur)θr−1)=Δ(F7r,+)(μ).\Delta_{U^r}(\operatorname{Ad}_{z_0}|_{U^r})=\Delta_{(\mathbb{F}_7^r,+)}(\theta_r(\operatorname{Ad}_{z_0}|_{U^r})\theta_r^{-1})=\Delta_{(\mathbb{F}_7^r,+)}(\mu).

Since (F7,+)(\mathbb{F}_7,+) is elementary abelian, its Frattini subgroup is {0}\{0\}, so

Δ(F7,+)(μ)=det⁡F7(μ).\Delta_{(\mathbb{F}_7,+)}(\mu)=\det_{\mathbb{F}_7}(\mu).

The matrix of μ\mu in the standard basis is

(d10⋱0dr).\begin{pmatrix} d_1 & 0\\ & \ddots\\ 0 & d_r \end{pmatrix}.

Consequently,

ΔUr(Ad⁡z0∣Ur)=det⁡F7(diag⁡(d1,…,dr))=∏i=1rdi=(∏i∈O2)(∏i∉O1)=2∣O∣.\Delta_{U^r}(\operatorname{Ad}_{z_0}|_{U^r})=\det_{\mathbb{F}_7}(\operatorname{diag}(d_1,\ldots,d_r))=\prod_{i=1}^{r}d_i=\left(\prod_{i\in O}2\right)\left(\prod_{i\notin O}1\right)=2^{|O|}.

We next compute the automorphism induced by Ad⁡z0\operatorname{Ad}_{z_0} on P0/UrP_0/U^r. Every element of P0=Ur⋊VP_0=U^r\rtimes V has the form xvxv, with x∈Urx\in U^r and v∈Vv\in V. Since z0z_0 centralizes VV and normalizes UrU^r,

z0(xv)z0−1=(z0xz0−1)(z0vz0−1)=(z0xz0−1)v,z0xz0−1∈Ur.z_0(xv)z_0^{-1}=(z_0xz_0^{-1})(z_0vz_0^{-1})=(z_0xz_0^{-1})v,\qquad z_0xz_0^{-1}\in U^r.

Therefore

Ad⁡‾z0(xvUr)=z0(xv)z0−1Ur=vUr=xvUr.\overline{\operatorname{Ad}}_{z_0}(xvU^r)=z_0(xv)z_0^{-1}U^r=vU^r=xvU^r.

Hence

Ad⁡‾z0=id⁡P0/Ur,ΔP0/Ur(Ad⁡‾z0)=1.\overline{\operatorname{Ad}}_{z_0}=\operatorname{id}_{P_0/U^r},\qquad\Delta_{P_0/U^r}(\overline{\operatorname{Ad}}_{z_0})=1.

Applying Lemma 3.1(2) to the Ad⁡z0\operatorname{Ad}_{z_0}-invariant normal subgroup Ur⊴P0U^r\trianglelefteq P_0, we obtain

ΔP0(Ad⁡z0)=ΔUr(Ad⁡z0∣Ur)ΔP0/Ur(Ad⁡‾z0)=2∣O∣⋅1=27k.\Delta_{P_0}(\operatorname{Ad}_{z_0})=\Delta_{U^r}(\operatorname{Ad}_{z_0}|_{U^r})\Delta_{P_0/U^r}(\overline{\operatorname{Ad}}_{z_0})=2^{|O|}\cdot1=2^{7^k}.

Finally, 7k≡1(mod3)7^k\equiv1\pmod{3} and 23=12^3=1 in F7×\mathbb{F}_7^{\times}, so 27k=2(23)(7k−1)/3=22^{7^k}=2^{(2^3)^{(7^k-1)/3}}=2. Thus ΔP0(Ad⁡z0)=2\Delta_{P_0}(\operatorname{Ad}_{z_0})=2, as required. Finally, let P∈Syl⁡7(W)P\in\operatorname{Syl}_7(W) be the prescribed subgroup. Choose w∈Ww\in W such that P=wP0w−1P=wP_0w^{-1} and put z=wz0w−1z=wz_0w^{-1}. Since GL⁡3(2)r⊴W\operatorname{GL}_3(2)^r\trianglelefteq W, we have z∈GL⁡3(2)rz\in\operatorname{GL}_3(2)^r and ∣z∣=3|z|=3. Also,

zPz−1=wz0P0z0−1w−1=wP0w−1=P.zPz^{-1}=wz_0P_0z_0^{-1}w^{-1}=wP_0w^{-1}=P.

By Lemma 3.1(1),

ΔP(Ad⁡z)=ΔP0(Ad⁡z0)=2,\Delta_P(\operatorname{Ad}_z)=\Delta_{P_0}(\operatorname{Ad}_{z_0})=2,

as required. □\square

We are now ready to prove Proposition 2.3.

Proof of Proposition 2.3. Suppose that such a soluble normal subgroup RR exists, and identify (B,⋅)/R(B,\cdot)/R with GL⁡3(2)r⋊T\operatorname{GL}_3(2)^r\rtimes T. Let (P,+)(P,+) and (Q,+)(Q,+) be the normal additive Sylow 77- and 33-subgroups, respectively. Then (P,⋅)R/R(P,\cdot)R/R is a Sylow 77-subgroup of (B,⋅)/R(B,\cdot)/R. Apply Lemma 4.4 with (p,q)=(7,3)(p,q)=(7,3). Since 7−1=3⋅27-1=3\cdot2, the character ψR,7,3=χR,72\psi_{R,7,3}=\chi_{R,7}^2 is trivial on RR and induces

ψ‾R,7,3:(B,⋅)/R→{1,2,4}.\overline{\psi}_{R,7,3}:(B,\cdot)/R\to\{1,2,4\}.

By Lemma 5.1, there is an element z∈GL⁡3(2)r∩N(B,⋅)/R((P,⋅)R/R)z\in\operatorname{GL}_3(2)^r\cap N_{(B,\cdot)/R}((P,\cdot)R/R) with Δ(P,⋅)R/R(Ad⁡z)=2\Delta_{(P,\cdot)R/R}(\operatorname{Ad}_z)=2. Since GL⁡3(2)r\operatorname{GL}_3(2)^r is perfect, zz lies in the derived subgroup of (B,⋅)/R(B,\cdot)/R and ψ‾R,7,3(z)=1\overline{\psi}_{R,7,3}(z)=1. By Lemma 4.6, however,

1=ψ‾R,7,3(z)=Δ(P,⋅)R/R(Ad⁡z)2=22=41=\overline{\psi}_{R,7,3}(z)=\Delta_{(P,\cdot)R/R}(\operatorname{Ad}z)^2=2^2=4

in F7×\mathbf{F}_7^\times, a contradiction. □\square

Declaration on the use of artificial intelligence

Theorem A originated from the author’s attempt to prove that every finite skew brace with supersoluble additive group has soluble multiplicative group. The proof emerged from a continuous and extensive interaction with ChatGPT (Sol 6.1), lasting several months, during which the author guided the mathematical investigation through questions, discussions, and successive refinements of the arguments. A crucial turning point was the idea of using the iterated Frattini series of finite pp-groups to construct determinant characters, which ultimately provided the key tool for establishing the main result. The initial proof generated by ChatGPT was incomprehensible in its original form and could not be presented as a readable mathematical argument without substantial reconstruction. The author subsequently undertook an extensive effort to understand, reconstruct, reorganize, and rewrite the arguments from the ground up, with the aim of making every step of the reasoning rigorous, transparent, and accessible to the mathematical community. All necessary mathematical verifications were carried out by the author, who takes full responsibility for the correctness of the results and the final version of the manuscript.

References

Department of Mathematics and Computer Science, University of Cagliari, Italy

Email address: [email protected]

References

  1. [1]N. P. Byott, Uniqueness of Hopf Galois structure for separable field extensions, Comm. Algebra 24 (1996), no. 10, 3217–3228.
  2. [2]N. P. Byott, Solubility criteria for Hopf–Galois structures, New York J. Math. 21 (2015), 883–903.arxiv.org/abs/1412.5923
  3. [3]N. P. Byott, On insoluble transitive subgroups in the holomorph of a finite soluble group, J. Algebra 638 (2024), 1–31.arxiv.org/abs/2205.13464
  4. [4]S. Camp-Mora, J. Pérez-Alemany and P. Pérez-Altarriba, Idealisation of skew left braces, Mediterr. J. Math. 23 (2026), Article 197. doi:10.1007/s00009-026-03191-3.
  5. [5]F. Cedó, A. Smoktunowicz and L. Vendramin, Skew left braces of nilpotent type, Proc. Lond. Math. Soc. (3) 118 (2019), no. 6, 1367–1392.DOI
  6. [6]M. Damele, A rigidity theorem for skew braces with multiplicative group S × T, Int. J. Group Theory, to appear; arXiv:2603.17805.
  7. [7]M. Damele, Finite skew braces whose additive group is a Z-group, Arch. Math. (Basel), to appear; arXiv:2603.22980.
  8. [8]M. Damele, Simple skew braces with cyclic Sylow subgroups, preprint (2026); arXiv:2607.17125.arxiv.org/abs/2607.17125
  9. [9]M. Di Matteo, M. Ferrara and M. Trombetti, A counterexample to Byott’s conjecture for finite skew braces, preprint (2026), arXiv:2607.22795.arxiv.org/abs/2607.22795
  10. [10]P. Etingof, T. Schedler and A. Soloviev, Set-theoretical solutions to the quantum Yang–Baxter equation, Duke Math. J. 100 (1999), no. 2, 169–209.arxiv.org/abs/math/9801047
  11. [11]I. Gorshkov and T. Nasybullov, Finite skew braces with solvable additive group, J. Algebra 574 (2021), 172–183.arxiv.org/abs/2006.00466
  12. [12]C. Greither and B. Pareigis, Hopf Galois theory for separable field extensions, J. Algebra 106 (1987), 239–258.DOI
  13. [13]L. Guarnieri and L. Vendramin, Skew braces and the Yang–Baxter equation, Math. Comp. 86 (2017), no. 307, 2519–2534.DOI
  14. [14]B. Li, T. Nasybullov and V. Zadornov, Comparison of addition and multiplication in a skew brace, preprint (2025), arXiv:2511.23232.
  15. [15]T. Nasybullov, Connections between properties of the additive and the multiplicative groups of a two-sided skew brace, J. Algebra 540 (2019), 156–167.arxiv.org/abs/1809.09418
  16. [16]D. J. S. Robinson, A course in the theory of groups, 2nd ed., Graduate Texts in Mathematics, vol. 80, Springer-Verlag, New York, 1996.
  17. [17]W. Rump, Braces, radical rings, and the quantum Yang–Baxter equation, J. Algebra 307 (2007), no. 1, 153–170.DOI
  18. [18]A. Smoktunowicz and L. Vendramin, On skew braces (with an appendix by N. Byott and L. Vendramin), J. Comb. Algebra 2 (2018), no. 1, 47–86.DOI
  19. [19]C. Tsang, Non-abelian simple groups which occur as the type of a Hopf–Galois structure on a solvable extension, Bull. London Math. Soc. 55 (2023), no. 5, 2324–2340.DOI
  20. [20]C. Tsang and C. Qin, On the solvability of regular subgroups in the holomorph of a finite solvable group, Int. J. Algebra Comput. 30 (2020), no. 2, 253–265.DOI
  21. [21]L. Vendramin, Problems on skew left braces, Adv. Group Theory Appl. 7 (2019), 15–37.DOI

Paper details

vlnc:W_wf7a6fqv
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC0-1.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

ChatGPT (Sol 6.1)

View statement
Looking for related papers…
Contents