A nonstandard rational permutation representation PE3(Z)→GL⁡7(Q)PE_3(\mathbb Z)\to\operatorname{GL}_7(\mathbb Q)

10.43

Problem

Describe, at least in particular cases, homomorphisms from projective elementary groups that fail the standard Weyl-image condition. Can such a homomorphism be given explicitly over the integers?

ρ0:PE3(Z)⟶GL⁡Q(W).\rho_0:PE_3(\mathbb Z)\longrightarrow\operatorname{GL}_{\mathbb Q}(W).

The obstruction below compares nonidentity involutory Weyl images with signed Weyl matrices of order four. It is an explicit example, not a classification of all homomorphisms.

The representation and its obstruction

For a commutative ring RR, let En(R)≤GL⁡n(R)E_n(R)\leq\operatorname{GL}_n(R) be generated by the transvections tij(r)t_{ij}(r), and put

tij(r)=1+reij,aij=tij(1)tji(−1)tij(1)(i≠j, r∈R).t_{ij}(r)=1+r e_{ij},\qquad a_{ij}=t_{ij}(1)t_{ji}(-1)t_{ij}(1) \quad(i\ne j,\ r\in R).

We write PEn(R)PE_n(R) for the image of En(R)E_n(R) in GL⁡n(R)/Z(GL⁡n(R))\operatorname{GL}_n(R)/Z(\operatorname{GL}_n(R)), and xˉ\bar x for the projective image of x∈En(R)x\in E_n(R).

Problem 10.43 of the Kourovka Notebook (1) asks, at least in particular cases, for homomorphisms outside a Weyl-image condition. That condition identifies all images of aˉij\bar a_{ij} with signed Weyl matrices over an endomorphism ring, through an injective group homomorphism and a module isomorphism. The standard-description theory was developed by Petechuk (2); trivial Weyl-square images also occur in the analysis of Petechuk and Petechuk (3). The example below fails the printed condition for every permitted embedding.

Set Ω=F23∖{0}\Omega=\mathbb F_2^3\setminus\{0\} and W=QΩW=\mathbb Q^{\Omega}, with coordinate basis (bv)v∈Ω(b_v)_{v\in\Omega}. For x∈GL⁡3(Z)x\in\operatorname{GL}_3(\mathbb Z), write x~\widetilde x for its reduction modulo two and define

ρ(x)bv=bx~v(v∈Ω).(2) \tag{2} \rho(x)b_v=b_{\widetilde x v}\qquad(v\in\Omega).

For a rational vector space PP, put DP=End⁡Q(P)D_P=\operatorname{End}_{\mathbb Q}(P), with 1DP1_{D_P} the identity of PP, and define

aP∗=(010−100001)∈GL⁡3(DP).(3) \tag{3} a_P^*=\begin{pmatrix}0&1&0\\-1&0&0\\0&0&1\end{pmatrix} \in\operatorname{GL}_3(D_P).

Theorem 1. The representation (2) induces a homomorphism

ρ0:PE3(Z)⟶GL⁡Q(W),ρ0(xˉ)=ρ(x),\rho_0:PE_3(\mathbb Z)\longrightarrow\operatorname{GL}_{\mathbb Q}(W), \qquad \rho_0(\bar x)=\rho(x),

where ∣Ω∣=7|\Omega|=7. For every i≠ji\ne j,

ρ0(aˉij)2=1,ρ0(aˉij)≠1.(5) \tag{5} \rho_0(\bar a_{ij})^2=1,\qquad \rho_0(\bar a_{ij})\ne1.

For any rational vector spaces P,TP,T, any linear isomorphism g:W→P3⊕Tg:W\to P^3\oplus T, and any injective homomorphism

τ:GL⁡3(DP)⟶GL⁡Q(P3⊕T),\tau:\operatorname{GL}_3(D_P)\longrightarrow \operatorname{GL}_{\mathbb Q}(P^3\oplus T),

one has

gρ0(aˉ12)g−1≠τ(aP∗).(7) \tag{7} g\rho_0(\bar a_{12})g^{-1}\ne\tau(a_P^*).

Consequently ρ0\rho_0 fails the condition in Problem 10.43 and is not induced by a standard homomorphism.

No finiteness, freeness, or nonzero hypothesis is imposed on PP or TT. The theorem concerns this particular representation.

Descent and Weyl images

Reduction modulo two sends invertible matrices to invertible matrices and preserves products. Thus each x~\widetilde x permutes Ω\Omega. The convention in (2) gives

ρ(x)ρ(y)bv=bx~y~v=ρ(xy)bv,\rho(x)\rho(y)b_v=b_{\widetilde x\widetilde y v} =\rho(xy)b_v,

so ρ\rho is a homomorphism. There are 23−1=72^3-1=7 basis vectors.

If z=(zkl)z=(z_{kl}) belongs to the centre of GL⁡3(Z)\operatorname{GL}_3(\mathbb Z), then

z(1+eij)=(1+eij)z,zkiδjl=δkizjl(i≠j).z(1+e_{ij})=(1+e_{ij})z, \qquad z_{ki}\delta_{jl}=\delta_{ki}z_{jl}\quad(i\ne j).

Taking l=jl=j and k≠ik\ne i gives zki=0z_{ki}=0; taking k=ik=i and l=jl=j gives zii=zjjz_{ii}=z_{jj}. Thus z=cI3z=cI_3 for some c∈Zc\in\mathbb Z. Its reduction c~I3\widetilde cI_3 is invertible, so c~≠0\widetilde c\ne0 in F2\mathbb F_2 and hence c~=1\widetilde c=1. Therefore ρ(z)=1\rho(z)=1, and the restriction of ρ\rho to E3(Z)E_3(\mathbb Z) descends to its image in the quotient by the ambient centre, giving ρ0\rho_0.

For i≠ji\ne j, direct multiplication of the three defining transvections gives

a~ij 2=I3,a~ijei=ej.\widetilde a_{ij}^{\,2}=I_3,\qquad \widetilde a_{ij}e_i=e_j.

The first identity implies ρ(aij)2=1\rho(a_{ij})^2=1; the second gives ρ(aij)bei=bej≠bei\rho(a_{ij})b_{e_i}=b_{e_j}\ne b_{e_i}. This proves (5) for all six ordered pairs.

Failure of the standard formula

Lemma 2. Let PP be a rational vector space, HH a group, and τ:GL⁡3(DP)→H\tau:\operatorname{GL}_3(D_P)\to H an injective homomorphism. If y∈Hy\in H satisfies y2=1y^2=1 and y≠1y\ne1, then τ(aP∗)≠y\tau(a_P^*)\ne y.

Proof. Suppose τ(aP∗)=y\tau(a_P^*)=y. Injectivity gives (aP∗)2=1(a_P^*)^2=1. The (1,1)(1,1)-entry of (aP∗)2(a_P^*)^2 is −1DP-1_{D_P}, so −1DP=1DP-1_{D_P}=1_{D_P}. For each p∈Pp\in P, this implies 2p=02p=0. Multiplication by 22 is invertible on PP, and therefore P=0P=0. In that case aP∗=1a_P^*=1, which gives y=1y=1, a contradiction. ◻

Apply Lemma 2 with H=GL⁡Q(P3⊕T)H=\operatorname{GL}_{\mathbb Q}(P^3\oplus T) and y=gρ0(aˉ12)g−1y=g\rho_0(\bar a_{12})g^{-1}. Conjugation preserves the two properties in (5), and (7) follows.

To exclude the standard formula, factor its unital coefficient homomorphism through its image as Z→δB→ιDP\mathbb Z\xrightarrow{\delta}B\xrightarrow{\iota}D_P, with δ\delta surjective and ι\iota injective. Let f∈Bf\in B be the central idempotent and ν:B→Bop\nu:B\to B^{\mathrm{op}} the antiisomorphism. Writing ν^(b)\widehat\nu(b) for the underlying element of BB corresponding to ν(b)\nu(b), the coefficient matrix has entries

M(x)ij=ι ⁣(δ(xij)f+ν^(δ((x−1)ji))(1−f)).(11) \tag{11} M(x)_{ij}=\iota\!\left( \delta(x_{ij})f+ \widehat\nu\bigl(\delta((x^{-1})_{ji})\bigr)(1-f) \right).

All the coefficient maps preserve 0,1,−10,1,-1. For x=a12x=a_{12}, one has

x=(010−100001),x−1=(0−10100001).x=\begin{pmatrix}0&1&0\\-1&0&0\\0&0&1\end{pmatrix}, \qquad x^{-1}=\begin{pmatrix}0&-1&0\\1&0&0\\0&0&1\end{pmatrix}.

Evaluating (11) gives M(a12)=aP∗M(a_{12})=a_P^*: on each nonzero entry the two branches recombine using f+(1−f)=1f+(1-f)=1. Standardness would imply gρ0(aˉ12)g−1=τ(aP∗)g\rho_0(\bar a_{12})g^{-1}=\tau(a_P^*), contrary to (7). This completes the proof of Theorem 1.

References

Preprint · Lean (GitHub)

  1. E. I. Khukhro and V. D. Mazurov (eds.), The Kourovka Notebook: Unsolved problems in group theory, 21st ed., September 2026 update, Problem 10.43, arXiv:1401.0300v46.
  1. V. M. Petechuk, Homomorphisms of linear groups over rings, Mathematical Notes 45 (1989), 144–151, doi:10.1007/BF01158061.
  1. V. M. Petechuk and Yu. V. Petechuk, Images of formal matrices of elements of matrix groups over associative rings, Scientific Bulletin of Uzhhorod University. Series of Mathematics and Informatics 36(1) (2020), 16–29, doi:10.24144/2616-7700.2020.1(36).16-29.