betaFeedback

WorksMathematics

On four problems from the Kourovka Notebook: spectra, periodic automorphism groups, commutators, and operator simplicity

Abstract

We examine Problems 18.62(a), 18.63, 18.65, and 20.120 of the Kourovka Notebook. First, a family of semidirect products involving PSL⁡2(p2)\operatorname{PSL}_2(p^2) gives arbitrarily many pairwise nonisomorphic groups critical with respect to one simple-group spectrum, subject to explicit arithmetic congruences. Second, a completed-algebra construction using sparse torsion identities and a matrix Golod-Shafarevich-Vinberg inequality produces an infinite periodic group generated by two fixed-point-free automorphisms of order five of an abelian group. Third, we give a short matrix proof that the class of elements of order three in PSL⁡2(8)\operatorname{PSL}_2(8) contradicts the assertion in Problem 18.65 as written. Finally, we spell out how P. Hall's 1962 wreath-power theorems answer Problem 20.120 under the usual normal-subgroup meaning of simplicity for a group with one distinguished automorphism. The four arguments have substantially different provenance. In particular, the last is a consequence of a known construction, not a priority claim. The two longer constructions are recorded in detail for independent mathematical scrutiny.

Statements, conventions, and limits of the claims

For a finite group XX let ω(X)\omega(X) be the set of orders of its elements. A section of XX is H/NH/N, where N⊴H≤XN \trianglelefteq H \le X. A group is ω\omega-critical if its spectrum is ω\omega and no proper section has this spectrum. We count groups up to isomorphism. The four problems considered here are independent: Problem 18.62(a) asks for a universal bound on the number of ω(L)\omega(L)-critical groups as LL ranges over finite simple groups; 18.63 concerns two fixed-point-free order-five automorphisms of an abelian group generating a periodic group; 18.65 asserts a normal pp-radical conclusion for a class of pp-elements whose pairwise commutators are pp-elements, with p≠5p \ne5; and 20.120 concerns operator simplicity of a locally nilpotent group [1]. No claim is made here to solve the classification problem 18.62(b).

Theorem 1.1 (18.62(a)). For every positive integer NN there exist an odd prime pp and NN pairwise nonisomorphic ω(PSp⁡4(p))\omega(\operatorname{PSp}_4(p))-critical finite groups. Consequently there is no universal upper bound as in 18.62(a).

Theorem 1.2 (18.63). There exist an abelian group AA and fixed-point-free automorphisms α,β∈Aut⁡(A)\alpha, \beta\in\operatorname{Aut}(A), each of order five, such that ⟨α,β⟩\langle\alpha,\beta\rangle is infinite and periodic.

Theorem 1.3 (18.65). Let G=PSL⁡2(8)G = \operatorname{PSL}_{2}(8), p=3p = 3, and CC be its conjugacy class of elements of order three. Every commutator [c,d][c,d], c,d∈Cc,d \in C, has order one or nine, whereas O3(G)=1O_{3}(G) = 1. Hence the assertion of 18.65, as stated with “pp-element”, is false.

Theorem 1.4 (20.120; consequence of Hall). For any prime pp there is a countable, locally finite pp-group DD and an automorphism σ\sigma of infinite order such that DD is perfect, every normal σ\sigma-invariant subgroup of DD is 11 or DD, and CD(σm)=1C_{D}(\sigma^{m}) = 1 for every nonzero integer mm.

Critical groups for the spectrum of PSp4(p)

The arithmetic choice of prime

Fix N≥1N \ge1 and set B=2N+1B = 2N + 1. Let ΠB\Pi_{B} contain every odd prime factor of s2+1s^{2} + 1 for odd 1≤s≤B1 \le s \le B, as well as every odd prime factor of s2−1s^{2} - 1 for odd 3≤s≤B3 \le s \le B. Set

MB=4∏ℓ∈ΠBℓM_{B} = 4 \prod_{\ell\in\Pi_{B}} \ell

Dirichlet’s theorem yields a prime p>B+2p > B + 2 with p≡1(modMB)p \equiv1 \pmod{M_{B}}. The following elementary estimate is the reason for this choice.

Lemma 2.1. For s∈{−B,−B+2,…,B}s \in\{-B, -B + 2, \ldots, B\} and ε∈{±1}\varepsilon\in\{\pm1\}, put u=ε+psu = \varepsilon+ p^{s}. Then

gcd⁡(p2+1,u)=2,\gcd(p^{2} + 1,u) = 2,

and the integer gcd⁡(p2−1,u)/2\gcd(p^{2} - 1,u)/2 divides either p−1p - 1 or p+1p + 1.

Proof. If an odd prime ℓ\ell divides p2+1p^{2} + 1 and uu, then p2≡−1p^{2} \equiv-1 and ps≡−ε(modℓ)p^{s} \equiv-\varepsilon\pmod{\ell}, whence ℓ∣s2+1\ell\mid s^{2} + 1. Our choice gives p≡1(modℓ)p \equiv1 \pmod{\ell}, a contradiction. Now v2(p2+1)=1v_{2}(p^{2} + 1) = 1 and uu is even, proving the first assertion.

If an odd prime ℓ\ell divides p2−1p^{2} - 1 and uu, then p≡±1(modℓ)p \equiv\pm1 \pmod{\ell}. The case p≡−1p \equiv-1 implies s≡ε(modℓ)s \equiv\varepsilon\pmod{\ell}. Unless s=εs = \varepsilon, this forces ℓ∣s2−1\ell\mid s^{2} - 1 (the case s=−εs = -\varepsilon produces only the factor two), contradicting p≡1(modℓ)p \equiv1 \pmod{\ell}. Thus, when s≠εs \ne\varepsilon, the odd part of the gcd divides p−1p - 1. Since p≡1(mod4)p \equiv1 \pmod{4}, v2(p+1)=1v_{2}(p + 1) = 1; after dividing the gcd by two, its remaining 22-part also divides p−1p - 1. When s=εs = \varepsilon, u=ε(p+1)u = \varepsilon(p + 1) and the gcd is p+1p + 1. □

Modules and element orders

Let

E=Fp2,S=PSL⁡2(E),H=S⋊⟨τ⟩,E = \mathbb{F}_{p^{2}}, \qquad S = \operatorname{PSL}_{2}(E), \qquad H = S \rtimes\langle\tau\rangle,

where τ\tau is the field automorphism x↦xpx \mapsto x^{p} of order two. For any odd b∈{3,5,…,B}b \in\{3,5,\ldots,B\} put

Wb=U⊗ESym⁡b(U)(p),W_{b} = U \otimes_{E} \operatorname{Sym}^{b}(U)^{(p)},

with UU the natural two-dimensional module for SL⁡2(E)\operatorname{SL}_{2}(E). Here the superscript (p)(p) denotes precomposition with the entrywise pp-power automorphism. Since b<pb < p and bb is odd, Steinberg’s tensor-product classification [3 Lemma 6] identifies WbW_b as a simple ESES-module: the central involution of SL2(E)\mathrm{SL}_2(E) acts by (−1)1+b=1(-1)^{1+b}=1. Set Vb=Res⁡E/FpWbV_b=\operatorname{Res}_{E/\mathbb{F}_p} W_b. Entrywise Frobenius on the usual tensor basis defines a Fp\mathbb{F}_p-linear involution FF satisfying Fρ(g)F−1=ρ(τ(g))F\rho(g)F^{-1}=\rho(\tau(g)), hence extends VbV_b to an FpH\mathbb{F}_pH-module. Over an algebraic closure,

Vb⊗FpF‾p≅W(1,b)⊕W(b,1).V_b\otimes_{\mathbb{F}_p}\overline{\mathbb{F}}_p\cong W(1,b)\oplus W(b,1).

The two constituents are nonisomorphic because b>1b>1, and the outer involution interchanges them. It follows that VbV_b is irreducible as an FpH\mathbb{F}_pH-module and

dim⁡FpVb=4(b+1).(1)\dim_{\mathbb{F}_p}V_b=4(b+1). \tag*{(1)}

The representation is faithful: the nontrivial simple group SS acts nontrivially, and τ\tau interchanges its two constituents.

We use the following standard element-order formulas, recorded in [3 Lemmas 3 and 5], for p>3p>3:

ω(PSp4(p))=Div⁡{p(p−1), p(p+1), p2−12, p2+12},(2)\omega(\mathrm{PSp}_4(p))=\operatorname{Div}\left\{p(p-1),\ p(p+1),\ \frac{p^2-1}{2},\ \frac{p^2+1}{2}\right\}, \tag*{(2)}

where Div⁡\operatorname{Div} denotes the set of positive divisors, and

ω(H)=ω(S)∪2ω(PSL2(p))⊆ω(PSp4(p)).(3)\omega(H)=\omega(S)\cup2\omega(\mathrm{PSL}_2(p))\subseteq\omega(\mathrm{PSp}_4(p)). \tag*{(3)}

Lemma 2.2 (Orders in the field-automorphism coset). Every p′p'-element of H∖SH\setminus S has order dividing p−1p-1 or p+1p+1, and both orders p−1p-1 and p+1p+1 occur in this coset. Every element of this coset whose order is divisible by pp has order 2p2p.

Proof. Write h=gτ‾h=g\overline{\tau} with g∈SL2(E)g\in\mathrm{SL}_2(E), and put n=gg(p)n=g g^{(p)}, where (p)(p) denotes entrywise Frobenius. Then h2=n‾h^2=\overline{n} and

tr⁡(n)p=tr⁡(g(p)g)=tr⁡(gg(p))=tr⁡(n).\operatorname{tr}(n)^p=\operatorname{tr}(g^{(p)}g)=\operatorname{tr}(g g^{(p)})=\operatorname{tr}(n).

Thus tr⁡(n)∈Fp\operatorname{tr}(n)\in\mathbb{F}_p and det⁡(n)=1\det(n)=1. If nn is semisimple, its eigenvalues either belong to Fp×\mathbb{F}_p^\times or form a Frobenius-conjugate pair of norm one in E×E^\times. Hence r=ord⁡(n)r=\operatorname{ord}(n) divides p−1p-1 or p+1p+1. Since hh is outside the subgroup of index two,

ord⁡(h)=2ord⁡(n‾)=2rgcd⁡(r,2).\operatorname{ord}(h)=2\operatorname{ord}(\overline{n})=\frac{2r}{\gcd(r,2)}.

If rr is even, this equals rr; if rr is odd, 2r2r still divides the corresponding even integer p−1p-1 or p+1p+1. If nn is not semisimple, its projective image is a nonidentity unipotent element of order pp, and therefore ord⁡(h)=2p\operatorname{ord}(h)=2p.

To realize p−1p-1, choose λ∈Fp×\lambda\in\mathbb{F}_p^\times of order p−1p-1 and x∈E×x\in E^\times with xxp=λxx^p=\lambda. Such xx exists by surjectivity of the norm map. For g=diag⁡(x,x−1)g=\operatorname{diag}(x,x^{-1}) we have gg(p)=diag⁡(λ,λ−1)g g^{(p)}=\operatorname{diag}(\lambda,\lambda^{-1}), so gτ‾g\overline{\tau} has order p−1p-1. To realize p+1p+1, choose λ∈E×\lambda\in E^\times of order p+1p+1 and put t=λ+λ−1∈Fpt=\lambda+\lambda^{-1}\in\mathbb{F}_p. Since λ≠−1\lambda\ne-1, we have t+2≠0t+2\ne0. Choose x∈E×x\in E^\times with xxp=t+2xx^p=t+2 and set

g=(0−11x).g=\begin{pmatrix}0&-1\\1&x\end{pmatrix}.

Then det⁡(g)=1\det(g)=1 and tr⁡(gg(p))=xxp−2=t\operatorname{tr}(gg^{(p)})=xx^p-2=t. Thus gg(p)gg^{(p)} has eigenvalues λ,λ−1\lambda,\lambda^{-1} and order p+1p+1, so gˉτ\bar{g}\tau has order p+1p+1.

Lemma 2.3. For every indicated bb, the semidirect product Gb=Vb⋊HG_b=V_b\rtimes H has the same spectrum as PSp⁡4(p)\operatorname{PSp}_4(p).

Proof. For a semisimple lift h~∈SL⁡2(E)\tilde{h}\in\operatorname{SL}_2(E) with eigenvalues λ,λ−1\lambda,\lambda^{-1}, the weights on W(1,b)W(1,b) are

λε+ps,ε∈{±1},s=−b,−b+2,…,b.(4)\lambda^{\varepsilon+ps},\qquad\varepsilon\in\{\pm1\},\qquad s=-b,-b+2,\ldots,b. \tag*{(4)}

On W(b,1)W(b,1) they are λs+pε\lambda^{s+p\varepsilon}. Because p2≡−1(modp2+1)p^2\equiv-1\pmod{p^2+1}, multiplication by pp transforms the latter exponents into exponents of the former shape up to signs. Lemma 2.1 shows that a nonidentity semisimple element of a nonsplit torus of SS has no fixed vector on VbV_b; for a semisimple h∈Sh\in S in a split torus,

CVb(h)≠0⟹ord⁡(h)∣p−1 or ord⁡(h)∣p+1.(5)C_{V_b}(h)\ne0\quad\Longrightarrow\quad\operatorname{ord}(h)\mid p-1\ \text{or}\ \operatorname{ord}(h)\mid p+1. \tag*{(5)}

Indeed, eigenvalues in a split (respectively nonsplit) torus have orders dividing p2−1p^2-1 (respectively p2+1p^2+1), while the image in SS identifies the central signs. For a unipotent u∈Su\in S of order pp, the nilpotence index of ρ(u)−1\rho(u)-1 on WbW_b is at most b+2<pb+2<p. Consequently

1+ρ(u)+⋯+ρ(u)p−1=(ρ(u)−1)p−1=0.1+\rho(u)+\cdots+\rho(u)^{p-1}=(\rho(u)-1)^{p-1}=0.

By Lemma 2.2, elements of H∖SH\setminus S with nontrivial pp-part have order 2p2p; their square has order pp, and the same geometric sum argument gives 1+h+⋯+h2p−1=01+h+\cdots+h^{2p-1}=0 as an endomorphism of VbV_b. Thus the extension introduces no elements of order p2p^2 or 2p22p^2.

For a p′p'-element h∈Hh\in H of order mm, the only possible orders above hh in Vb⋊HV_b\rtimes H are mm and pmpm, and the second is possible precisely when CVb(h)≠0C_{V_b}(h)\ne0. If h∈Sh\in S, equation (5) puts pmpm in (2); if h∉Sh\notin S, Lemma 2.2 shows that m∣p−1m\mid p-1 or m∣p+1m\mid p+1. We have proved ω(Gb)⊆ω(PSp⁡4(p))\omega(G_b)\subseteq\omega(\operatorname{PSp}_4(p)).

The maximal orders (p2±1)/2(p^2\pm1)/2 are already represented in SS. By Lemma 2.2, choose t−,t+∈H∖St_-,t_+\in H\setminus S of orders p−1p-1 and p+1p+1, respectively. Their squares are in SS, of orders (p−1)/2(p-1)/2 and (p+1)/2(p+1)/2. For t−2t_-^2 one may take an eigenvalue λ\lambda of order p−1p-1 with λp=λ\lambda^p=\lambda; the weight 1−p1-p in (4) is then trivial. For t+2t_+^2 one may take λ\lambda of order p+1p+1, with λp=λ−1\lambda^p=\lambda^{-1}; the weight 1+p1+p is trivial. The involution τ\tau exchanges the two absolutely simple constituents. Thus from a nonzero vector ww fixed by t±2t_\pm^2 in one constituent, w+t±ww+t_\pm w is a nonzero vector fixed by t±t_\pm. Therefore the semidirect product contains elements of orders p(p−1)p(p-1) and p(p+1)p(p+1). All four maximal element orders in (2) occur, proving equality.

Criticality

Lemma 2.4. Each GbG_b is ω(PSp4(p))\omega(\mathrm{PSp}_4(p))-critical.

Proof. Write V=VbV=V_b and G=V⋊HG=V\rtimes H. Suppose a subgroup M≤GM\leq G has the full spectrum. Its projection to HH contains elements of orders

A=p2−12,B′=p2+12.A=\frac{p^2-1}{2},\qquad B'=\frac{p^2+1}{2}.

Both elements project to SS, because they are p′p'-elements and Lemma 2.2 excludes both orders in the outer coset: both AA and B′B' exceed p+1p+1. By the classification of maximal subgroups of PSL2(p2)\mathrm{PSL}_2(p^2) (Dickson’s theorem [2 Chapter XII]), a proper subgroup cannot contain both torus orders AA and B′B' when p>5p>5: a Borel subgroup or a split torus normalizer misses B′B', a nonsplit torus normalizer misses AA, a subfield group misses B′B', and an exceptional subgroup A4A_4, S4S_4, or A5A_5 has insufficient order. Consequently S≤π(M)S\leq\pi(M), and either π(M)=S\pi(M)=S or π(M)=H\pi(M)=H.

If π(M)=S\pi(M)=S, then M≤V⋊SM\leq V\rtimes S. But this group does not have elements of order p(p−1)p(p-1) or p(p+1)p(p+1). Indeed a semisimple element of SS of order p−1p-1 (resp. p+1p+1) lifts to eigenvalues in SL2(E)\mathrm{SL}_2(E) of order 2(p−1)2(p-1) (resp. 2(p+1)2(p+1)). To fix a weight in (4), the exponent ε+ps\varepsilon+p^s would have to be divisible by the corresponding order. Reducing modulo p∓1p\mp1 gives s=∓εs=\mp\varepsilon since ∣s∣≤B<p−2|s|\leq B<p-2; the exponent is then only ±(p∓1)\pm(p\mp1) and fails the required divisibility by 2(p∓1)2(p\mp1). The same holds for the Galois-conjugate constituent. Thus π(M)≠S\pi(M)\neq S.

If π(M)=H\pi(M)=H, then M∩VM\cap V is HH-stable, hence either 00 or VV by irreducibility. In the first case M≅HM\cong H and cannot have the full spectrum; in the second M=GM=G. Hence no proper subgroup has the full spectrum.

Finally, since VV is faithful and irreducible, every nontrivial normal subgroup NN of GG contains VV: if N∩V=1N\cap V=1, then NN centralizes VV and lies in CG(V)=VC_G(V)=V, a contradiction. The quotient G/V≅HG/V\cong H does not have the full spectrum. Therefore no nontrivial quotient does either. Since a section K/NK/N has its spectrum contained in ω(K)\omega(K), the two facts prove criticality. □

Proof of Theorem 1.1. For each NN, choose pp as in Lemma 2.1, and take the NN modules indexed by b=3,5,…,2N+1b=3,5,\ldots,2N+1. Lemmas 2.3 and 2.4 give the desired critical groups. They are pairwise nonisomorphic because dim⁡FpVb=4(b+1)\dim_{\mathbb{F}_p}V_b=4(b+1) are distinct, and VbV_b is the unique minimal nontrivial normal subgroup of GbG_b. □

Two fixed-point-free automorphisms of order five

Cyclotomic algebra and its formal neighbourhood

Put K=F11K=\mathbb{F}_{11} and let

U=K⟨a,b⟩/(Φ5(a),Φ5(b)),Φ5(X)=1+X+X2+X3+X4.(6)U=K\langle a,b\rangle/(\Phi_5(a),\Phi_5(b)),\qquad\Phi_5(X)=1+X+X^2+X^3+X^4. \tag*{(6)}

Since Φ5\Phi_5 splits into four distinct linear factors over KK,

U≅K4∗KK4.U\cong K^4 *_K K^4.

In UU, both a,ba,b have fifth power one; moreover

(X−1)(X3+2X2+3X+4)=Φ5(X)−5(X-1)(X^3+2X^2+3X+4)=\Phi_5(X)-5

shows that a−1,b−1a-1,b-1 are units (as 5∈K×5\in K^\times). Let L=ker⁡(U→Uab)L=\ker(U\to U_{\mathrm{ab}}) be its commutator ideal. Then S0=U/L≅K16S_0=U/L\cong K^{16}. We use m\mathfrak{m} for the arrow ideal of a completed path algebra.

Lemma 3.1 (Formal neighbourhood). There is an isomorphism of topological KK-algebras

U^L=lim←⁡U/Ln≅KQ^,\widehat{U}_L=\varprojlim U/L^n\cong\widehat{KQ},

where the vertices of QQ are pairs (i,j)(i,j) with 1≤i,j≤41\leq i,j\leq4, and one arrow runs from (i,j)(i,j) to (k,l)(k,l) exactly when i≠ki\ne k and j≠lj\ne l. Every vertex has nine incoming and nine outgoing arrows.

Proof. Write M=L/L2M=L/L^2. The square-zero quotient U/L2U/L^2 splits over the separable algebra S0S_0. Fix such a splitting and write exe_x for the primitive idempotent corresponding to a character xx, and KxK_x for the associated one-dimensional left UU-module. Every extension of KxK_x by KyK_y is annihilated by L2L^2. The square-zero description gives

Ext⁡U1(Kx,Ky)≅Hom⁡K(eyMex,K).\operatorname{Ext}^1_U(K_x,K_y)\cong\operatorname{Hom}_K(e_yMe_x,K).

Indeed, after splitting the extension as an S0S_0-module, the remaining action is precisely a linear map from eyMexe_yMe_x to Hom⁡K(Kx,Ky)\operatorname{Hom}_K(K_x,K_y). We can therefore compute the corner dimensions of MM from upper triangular extensions, modulo changes of splitting. A character of S0S_0 is specified by a choice of nontrivial fifth root of unity for each of a,ba,b. For two distinct characters x=(i,j)x=(i,j) and y=(k,l)y=(k,l), a square-zero upper triangular extension is determined by the off-diagonal entries of a,ba,b. The first entry is free exactly when i≠ki\ne k, the second exactly when j≠lj\ne l. For distinct characters, we quotient out the one-dimensional space of inner derivations arising from changes of splitting. For equal characters both entries must be zero. Thus the dimensions of the off-diagonal components of MM are one if both labels differ, and zero otherwise; the diagonal components are zero. This identifies TS0(M)T_{S_0}(M) with the ordinary path algebra KQKQ.

For completeness we justify the passage to completions. A homomorphism Kr→E/JK^r\to E/J with nilpotent JJ lifts to EE by lifting a complete set of orthogonal idempotents: for a square-zero ideal and x2−x=n∈Jx^2-x=n\in J one corrects xx to x+(1−2x)nx+(1-2x)n; successive lifting handles higher nilpotence. Therefore U=K4∗KK4U=K^4*_K K^4 has the lifting property for nilpotent algebra extensions. The algebra TS0(M)T_{S_0}(M) has the same property, since S0≅K16S_0\cong K^{16} and its enveloping algebra S0⊗KS0opS_0\otimes_K S_0^{\mathrm{op}} are semisimple, so the S0S_0-bimodule MM is projective.

Put T=TS0(M)T=T_{S_0}(M), with arrow ideal m\mathfrak{m}. The splitting of U/L2U/L^2 gives an isomorphism

f2:T/m2⟶U/L2f_2:T/\mathfrak{m}^2\longrightarrow U/L^2

inducing the prescribed identifications in degrees zero and one. Suppose inductively that a compatible isomorphism fn:T/mn→U/Lnf_n:T/\mathfrak{m}^n\to U/L^n has been constructed, where n≥2n \ge2. Compose fnf_n with the quotient map from the untruncated algebra TT. Formal smoothness of TT lifts this homomorphism to f~n+1:T→U/Ln+1\widetilde{f}_{n+1}: T \to U/L^{n+1}. The image of m\mathfrak{m} lies in L/Ln+1L/L^{n+1}, because its reduction modulo LnL^n does. Hence this lift annihilates mn+1\mathfrak{m}^{n+1} and induces a compatible map

fn+1:T/mn+1⟶U/Ln+1.f_{n+1}: T/\mathfrak{m}^{n+1} \longrightarrow U/L^{n+1}.

It is surjective: it is an isomorphism in degrees zero and one, and the higher associated graded pieces are generated by products of degree-one elements. For the reverse comparison, compose U→U/LnU \to U/L^n with fn−1f_n^{-1}. Formal smoothness of the untruncated algebra UU gives a lift U→T/mn+1U \to T/\mathfrak{m}^{n+1}. This sends LL into m/mn+1\mathfrak{m}/\mathfrak{m}^{n+1} and therefore factors through a map U/Ln+1→T/mn+1U/L^{n+1} \to T/\mathfrak{m}^{n+1}, which is surjective by the same associated graded argument. Both truncated algebras are finite-dimensional. For U/Ln+1U/L^{n+1}, this follows because U/LU/L is finite-dimensional and LL is generated as a two-sided ideal by ab−baab-ba; each layer is a quotient of a tensor power over S0S_0 of L/L2L/L^2. The two surjections force equality of dimensions, so fn+1f_{n+1} is an isomorphism. The forward lifts are compatible by construction. Passing to inverse limits proves U^L≅KQ^\widehat{U}_L \cong\widehat{KQ}. □

Sparse torsion relations

Enumerate the group words in a±1,b±1a^{\pm1},b^{\pm1} as w1,w2,…w_1,w_2,\ldots, allowing repetitions. Their images in U/LU/L have order dividing five, so un=wn5−1∈Lu_n=w_n^5-1\in L. Define

rn=wn5⋅11n−1=(wn5−1)11n∈L11n.(7)r_n=w_n^{5\cdot11^n}-1=(w_n^5-1)^{11^n}\in L^{11^n}. \tag*{(7)}

The equality uses the characteristic-11 Frobenius identity for powers of the single element wn5−1w_n^5-1; no commutativity of UU is required. Let

R^=U^L/(r1,r2,…)‾.\widehat{R}=\widehat{U}_L/\overline{(r_1,r_2,\ldots)}.

We use the matrix Golod–Shafarevich–Vinberg inequality for quotients of completed path algebras, in the following form proved for finite families of (possibly inhomogeneous) relations in [4 Theorem 1.2]: if the quotient is finite-dimensional, then for every 0<ρ<10<\rho<1 there is a vector H(ρ)H(\rho) with all entries at least one satisfying

1≤(I−ρMQT+R(ρ))H(ρ).(8)\mathbf{1}\leq(I-\rho M_Q^{\mathsf{T}}+R(\rho))H(\rho). \tag*{(8)}

Here MQM_Q is the adjacency matrix and R(ρ)R(\rho) is the matrix whose (t,s)(t,s)-entry sums ρdr\rho^{d_r} over relations with target tt, source ss and m\mathfrak{m}-adic order drd_r.

Lemma 3.2 (Finite extraction). Suppose JJ is the closed two-sided ideal generated by a possibly countable family R\mathcal{R} in a completed path algebra A=KQ^A=\widehat{KQ} of a finite quiver. If A/JA/J is finite-dimensional, there is a finite subfamily R0⊆R\mathcal{R}_0\subseteq\mathcal{R} such that A/(R0)A/(\mathcal{R}_0) is finite-dimensional.

Proof. Let m\mathfrak{m} be the arrow ideal. Finite-dimensionality implies md⊆J\mathfrak{m}^d\subseteq J for some dd (the image of m\mathfrak{m} is a nilpotent ideal in the finite-dimensional quotient). The vector space md/md+1\mathfrak{m}^{d}/\mathfrak{m}^{d+1} is finite-dimensional. As JJ is the closure of the ideal generated by R\mathcal{R}, we can choose finitely many relations giving

md⊆I0+md+1,I0=(R0‾).\mathfrak{m}^{d} \subseteq I_0+\mathfrak{m}^{d+1}, \qquad I_0=(\overline{\mathcal{R}_0}).

Pass to the complete algebra A/I0A/I_0 and let m‾\overline{\mathfrak{m}} be its arrow ideal. Then m‾d=m‾d+1\overline{\mathfrak{m}}^{d}=\overline{\mathfrak{m}}^{d+1}. The ideal m‾d\overline{\mathfrak{m}}^{d} is finitely generated as a right module by images of the finitely many paths of length dd, while m‾\overline{\mathfrak{m}} lies in the Jacobson radical (geometric series converge). Nakayama’s lemma gives m‾d=0\overline{\mathfrak{m}}^{d}=0. Thus A/I0A/I_0 is finite-dimensional. □

Proposition 3.3. The algebra R^\widehat{R} is infinite-dimensional over KK.

Proof. Suppose it is finite-dimensional. By Lemma 3.2, a finite subfamily of the rnr_n, after decomposition into vertex-to-vertex components, already gives a finite-dimensional quotient of KQ^K\widehat{Q}. Each such component has m\mathfrak{m}-adic order at least 11n11^n. For each fixed source vertex an rnr_n has at most 16 endpoint components. Therefore, for ρ=1/2\rho=1/2, every column sum of the relation matrix is bounded by

16∑n≥12−11n<164.16\sum_{n\geq1}2^{-11^n}<\frac{1}{64}.

Every vertex of QQ has indegree and outdegree nine. Hence every entry of the row vector

1T(I−12MQ^+R(1/2))\mathbf{1}^{\mathsf{T}}\left(I-\frac{1}{2}M\widehat{Q}+R(1/2)\right)

is at most 1−9/2+1/64<01-9/2+1/64<0. This is incompatible with (8) after multiplying on the left by 1T\mathbf{1}^{\mathsf{T}}, because all entries of H(1/2)H(1/2) are positive. Contradiction. □

Passing from a completed algebra to automorphisms

Proof of Theorem 1.2. Let BB be the image of the ordinary algebra UU in R^\widehat{R}. This image is dense. If BB were a finite-dimensional KK-algebra, it would be a finite set because K=F11K=\mathbb{F}_{11}, hence closed in the Hausdorff m\mathfrak{m}-adic topology. Density would imply B=R^B=\widehat{R}, contrary to Proposition 3.3. Thus BB is infinite-dimensional and contains a unit.

The images a,b∈Ba,b\in B are units of order exactly five: a5=b5=1a^5=b^5=1, and a−1,b−1a-1,b-1 are units, which precludes a=1a=1 or b=1b=1. Every element of the subgroup Γ=⟨a,b⟩≤B×\Gamma=\langle a,b\rangle\leq B^\times is represented by some word wnw_n and satisfies wn5⋅11n=1w_n^{5\cdot11^n}=1 by [7]. Hence Γ\Gamma is periodic. Since BB is generated as a KK-algebra by the units a,ba,b and their inverses are polynomials in them, we have B=K-span(Γ)B=K\text{-span}(\Gamma). If Γ\Gamma were finite then dim⁡KB<∞\dim_K B<\infty, a contradiction. Thus Γ\Gamma is infinite.

Take the additive group A=(B,+)A=(B,+) and let α=La\alpha=L_a, β=Lb\beta=L_b be left multiplications. They are automorphisms of exact order five and have no nonzero fixed points, since a−1,b−1a-1,b-1 are invertible in BB. The left regular map Γ→Aut⁡(A)\Gamma\to\operatorname{Aut}(A) is injective (evaluate on 11), so ⟨α,β⟩≅Γ\langle\alpha,\beta\rangle\cong\Gamma is infinite and periodic. □

A commutator counterexample in PSL⁡2(8)\operatorname{PSL}_2(8)

The condition investigated here is the one stated in the Kourovka Notebook; compare the related conjectures and commutator results of Guralnick–Malle and Guralnick–Tong-Viet–Tracey [5, 6].

We use [x,y]=x−1y−1xy[x,y]=x^{-1}y^{-1}xy. As F8\mathbb{F}_8 has even characteristic, G=PSL⁡2(8)=SL⁡2(8)G=\operatorname{PSL}_2(8)=\operatorname{SL}_2(8) is simple. The polynomial X2+X+1X^2+X+1 is irreducible over F8\mathbb{F}_8; therefore the trace-one matrices of determinant one constitute a single conjugacy class CC of cardinality 504/9=56504/9=56, and each has order three.

Proof of Theorem 1.3. For any c,d∈Cc,d\in C, put s=tr⁡(cd)∈F8s=\operatorname{tr}(cd)\in\mathbb{F}_8. The Fricke trace identity

tr⁡[c,d]=tr⁡(c)2+tr⁡(d)2+s2+tr⁡(c)tr⁡(d)s−2\operatorname{tr}[c,d]=\operatorname{tr}(c)^2+\operatorname{tr}(d)^2+s^2+\operatorname{tr}(c)\operatorname{tr}(d)s-2

gives, in characteristic two,

t:=tr⁡[c,d]=s2+s.(9)t:=\operatorname{tr}[c,d]=s^2+s. \tag*{(9)}

The additive trace Tr⁡F8/F2(t)\operatorname{Tr}_{\mathbb{F}_8/\mathbb{F}_2}(t) is zero:

t+t2+t4=s+s8=0.t+t^2+t^4=s+s^8=0.

Consequently t=0t=0 or tt is a root of

X3+X+1=0,(10)X^3+X+1=0, \tag*{(10)}

since the trace-zero subspace of F8\mathbb{F}_8 is precisely the set of roots of X(X3+X+1)X(X^3+X+1). If t=0t=0, then s2+s=0s^2+s=0, so s∈{0,1}s\in\{0,1\}. Lemma 4.1 below proves that cc and dd commute in these two cases. Thus [c,d]=1[c,d]=1.

If t≠0t\ne0, let z,z−1z,z^{-1} be the eigenvalues of [c,d][c,d] over F64\mathbb{F}_{64}. Then z+z−1=tz+z^{-1}=t. Equation (10) gives

z3+z−3=(z+z−1)3+(z+z−1)=t3+t=1.z^3+z^{-3}=(z+z^{-1})^3+(z+z^{-1})=t^3+t=1.

Thus z3z^3 is a nontrivial cube root of unity and zz has order nine. Hence [c,d][c,d] has order nine. Finally O3(G)=1O_3(G)=1 by simplicity, completing the proof. □

Lemma 4.1. If c,d∈SL⁡2(8)c,d\in\operatorname{SL}_2(8) have determinant one, trace one, and tr⁡(cd)∈{0,1}\operatorname{tr}(cd)\in\{0,1\}, then cd=dccd=dc.

Proof. Identify the natural two-dimensional F8\mathbb{F}_8-space with E=F64E=\mathbb{F}_{64} so that cc acts by multiplication by a primitive cube root ζ∈E\zeta\in E. Every F8\mathbb{F}_8-linear endomorphism of EE has a unique expression

d(x)=ux+vx8,u,v∈E.d(x)=ux+vx^8,\qquad u,v\in E.

Its trace and determinant over F8\mathbb{F}_8 are respectively u+u8u+u^8 and uu8+vv8uu^8+vv^8. In particular

tr⁡(d)=u+u8,tr⁡(cd)=ζu+ζ8u8,det⁡(d)=uu8+vv8.\operatorname{tr}(d)=u+u^8,\qquad\operatorname{tr}(cd)=\zeta u+\zeta^8u^8,\qquad\det(d)=uu^8+vv^8.

The first two equations constitute an invertible F8\mathbb{F}_8-linear system for u∈Eu\in E: its determinant is ζ+ζ8=1\zeta+\zeta^8=1, since ζ8=ζ2\zeta^8=\zeta^2. For (tr⁡(d),tr⁡(cd))=(1,1)(\operatorname{tr}(d),\operatorname{tr}(cd))=(1,1) the unique solution is u=ζu=\zeta; for (1,0)(1,0) it is u=ζ2u=\zeta^2. In either case uu8=1uu^8=1, so det(d)=1(d)=1 implies vv8=0vv^{8}=0, hence v=0v=0. Therefore dd is multiplication by ζ\zeta or ζ2\zeta^{2} and commutes with cc. □

Remark 4.2 (An independent finite certificate). A standard-library-only Python program enumerates all 504 matrices of SL2(8)\mathrm{SL}_{2}(8) and all 56256^{2} ordered pairs in CC. It finds exactly 112 identity commutators and 3024 commutators of order nine. The source is supplied as anc/check_psl2_8.py. The program checks a finite instance and is not used to justify any infinite construction.

Locally nilpotent operator-simple groups: Hall's construction

We spell out the convention used in Problem 20.120. A pair (D,σ)(D,\sigma) is operator-simple if D≠1D\ne1 and

N⊴D, σ(N)=N⟹N=1 or N=D.N\trianglelefteq D,\ \sigma(N)=N \quad\Longrightarrow\quad N=1\text{ or }N=D.

This is equivalent to DD being a minimal nontrivial normal subgroup of D⋊σZD\rtimes_{\sigma}\mathbb{Z}.

Let V=⨁n∈ZFpenV=\bigoplus_{n\in\mathbb{Z}}\mathbb{F}_{p}e_{n}, and for each nn define a permutation ana_{n} of VV by

an(v)={v+en,vj=0 for all j>n,v,otherwise.a_{n}(v)= \begin{cases} v+e_{n}, & v_{j}=0\text{ for all }j>n,\\ v, & \text{otherwise}. \end{cases}

Put W=⟨an:n∈Z⟩W=\langle a_{n}:n\in\mathbb{Z}\rangle and D=W′D=W'. For u≤vu\le v, the subgroup ⟨au,…,av⟩\langle a_{u},\ldots,a_{v}\rangle is the iterated regular wreath product of v−u+1v-u+1 copies of CpC_{p}. It is a finite pp-group. Thus WW is countable, locally finite, and locally nilpotent. It is nonabelian: a0a1(0)=e1a_{0}a_{1}(0)=e_{1} while a1a0(0)=e0+e1a_{1}a_{0}(0)=e_{0}+e_{1}. The shift s(en)=en+1s(e_{n})=e_{n+1} conjugates ana_{n} to an+1a_{n+1}; it induces an automorphism σ\sigma of DD.

We now apply precisely the two results of P. Hall [7 Theorems A and D] needed here:

(H1) for an ordered wreath power indexed by a set without a greatest element, its derived group is perfect and every normal subgroup of that derived group is normal in the whole wreath power;

(H2) if the order-preserving automorphism group of the indexing set acts irreducibly in Hall's sense, that derived group is a minimal nontrivial normal subgroup of the natural split extension.

Here the ordered set is Z\mathbb{Z}, without a greatest element; its order-preserving automorphism group consists of all translations, and is irreducible in Hall's sense. Both hypotheses therefore hold. Consequently D=W′=W′′D=W'=W'', and if N⊴DN\trianglelefteq D is invariant under σ\sigma, then (H1) gives N⊴WN\trianglelefteq W, while shift invariance makes it normal in W⋊ZW\rtimes\mathbb{Z}. By (H2), N=1N=1 or DD. The specialization to the integer-indexed wreath power and its extension by the shift is already discussed explicitly in [7 Section 1.7].

To check the extra fixed-point assertion, define for g∈Wg\in W its moving-coordinate set

M(g)={n∈Z:∃v∈V with (gv)n≠vn}.M(g)=\{n\in\mathbb{Z}:\exists v\in V\text{ with }(gv)_{n}\ne v_{n}\}.

Every g∈Wg \in W is a finite word in the ana_n, so M(g)M(g) is finite, and M(g)=∅M(g) = \varnothing only if g=1g = 1. Moreover M(σm(g))=M(g)+mM(\sigma^m(g)) = M(g) + m. If σm(g)=g\sigma^m(g) = g for m≠0m \ne0, this finite set is invariant under a nonzero translation and must be empty. Thus CW(σm)=1C_W(\sigma^m) = 1, hence CD(σm)=1C_D(\sigma^m) = 1. Since WW is nonabelian, D=W′≠1D = W' \ne1; the fixed-point assertion therefore also implies that σ\sigma has infinite order. This proves Theorem 1.4.

Remark 5.1 (Attribution and formulation). Hall’s construction predates the Kourovka Notebook question by decades. Theorem 1.4 is an observation about its operator-group interpretation; it must not be represented as an independently new group construction. If Problem 20.120 was intended to use a stronger meaning of “simple with an automorphism”, the statement would need to be revisited.

Verification record and scope

The short counterexample in Section 4 was checked in two independent ways: the trace calculation and the exhaustive finite-field program described in Remark 4.2. § 5 uses two specific theorems of Hall, whose content also appears in a published biographical account [8]. The two longer sections rely on standard results that have been explicitly identified: Steinberg’s tensor-product classification and the relevant spectrum formulas in [3], Dickson’s subgroup classification, Dirichlet’s theorem, and the completed-quiver inequality of [4]. The complete proofs of these external theorems are not reproduced here. The reader should particularly scrutinize the formal-neighbourhood and finite-extraction arguments in § 3 and the use of the outer field coset in § 2; these are the main technical interfaces of the submitted constructions.

AI assistance disclosure

An AI language model assisted in preparing the initial manuscript. The submitting author has disclosed the use of generative artificial intelligence and acknowledges responsibility for the mathematical claims and presentation. Novosibirsk State University, Novosibirsk, Russia

Email address: [email protected]

References

  1. [1]E. I. Khukhro and V. D. Mazurov (eds.), The Kourovka Notebook: Unsolved Problems in Group Theory, 21st edition (2026), Problems 18.62, 18.63, 18.65, and 20.120. November 2026 update. https://kourovkanotebook.org.wordpress.com/
  2. [2]L. E. Dickson, Linear Groups: With an Exposition of the Galois Field Theory, B. G. Teubner, Leipzig, 1901, Chapter XII. doi:10.5962/bhl.title.22174.DOI
  3. [3]Yu. V. Lytkin, On finite groups isospectral to the simple groups S₄(q), Siberian Electronic Mathematical Reports 15 (2018), 570–584. doi:10.17377/sm-2018-15-046.
  4. [4]W. Chang and Q. Tang, Obstructions to Jacobi-Finiteness of Quivers with Potentials, preprint, 2026, arXiv:2609.11200.arxiv.org/abs/2609.11200
  5. [5]R. Guralnick and G. Malle, Variations on the Baer–Suzuki theorem, Math. Z. 279 (2015), 981–1006. doi:10.1007/s00209-014-1399-y. Preprint: arXiv:1310.5909.DOI
  6. [6]R. M. Guralnick, H. P. Tong-Viet, and G. Tracey, Weakly subnormal subgroups and variations of the Baer-Suzuki theorem, J. London Math. Soc. 111 (2025), article e70057. doi:10.1112/jlms.70057.DOI
  7. [7]P. Hall, Wreath powers and characteristically simple groups, Proc. Cambridge Philos. Soc. 58 (1962), 170–184. doi:10.1017/S0305004100036379.DOI
  8. [8]J. E. Roseblade, Philip Hall, Bull. London Math. Soc. 16 (1984), 603–626, especially p. 623. Available at https://mathshistory.st-andrews.ac.uk/LMS/hall_lms_obit.pdf.

Paper details

vlnc:W_tykjhbun
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Mathematics

An AI language model

View statement
Looking for related papers…
Contents