Finite separating isotopes of Gm!+1G^{m!+1} and non-generation of isotope closures

9.4

Problem

If a variety, quasivariety, or pseudovariety M\mathcal M is generated by one finite group, is its isotope closure IMI\mathcal M generated by one finite quasigroup?

M=C(G),∣G∣<∞⟹?IM=C(A),∣A∣<∞.\mathcal M=\mathsf C(G),\quad |G|<\infty \quad\stackrel{?}{\Longrightarrow}\quad I\mathcal M=\mathsf C(A),\quad |A|<\infty.

The proof distinguishes modern pseudovarieties of finite algebras from the original convention using disjunctive identities.

Setup and result

A quasigroup is a nonempty algebra (A;⋅,\,/)(A;\cdot,\backslash,/) satisfying

x\(x⋅y)=y,x⋅(x\y)=y,(x⋅y)/y=x,(x/y)⋅y=x.(1) \tag{1} x\backslash(x\cdot y)=y,\quad x\cdot(x\backslash y)=y,\quad (x\cdot y)/y=x,\quad (x/y)\cdot y=x.

An isotopy from a quasigroup AA to a group HH consists of bijections α,β,γ:A→H\alpha,\beta,\gamma:A\to H such that

α(x)β(y)=γ(x⋅y)(x,y∈A).\alpha(x)\beta(y)=\gamma(x\cdot y)\qquad(x,y\in A).

For a class M\mathcal M of groups, let IMI\mathcal M be the class of all quasigroups isotopic to members of M\mathcal M.

Problem 9.4 of the Kourovka Notebook asks whether IMI\mathcal M is generated by a single finite quasigroup whenever M\mathcal M is generated by a single finite group (3). The question is posed for varieties, quasivarieties and pseudovarieties. The last term requires care: Gvaramiya’s original paper (2, p. 1048) uses classes defined by disjunctive identities, rather than pseudovarieties of finite algebras. We treat both conventions.

Write V(A)\mathsf{V}(A) and Q(A)\mathsf{Q}(A) for the classes defined by all identities and all quasiidentities, respectively, valid in AA. A quasiidentity has finitely many equational premises and one equational conclusion. For finite AA, write PV(A)\mathsf{PV}(A) for its closure under homomorphic images, subalgebras and finite direct products. All homomorphisms preserve the three operations in (1). For a group these operations are

x⋅y=xy,x\y=x−1y,x/y=xy−1.x\cdot y=xy,\qquad x\backslash y=x^{-1}y,\qquad x/y=xy^{-1}.

When the generating algebra is a group, the corresponding group class is understood within groups.

Theorem 1. Let GG be a nontrivial finite group. For every finite quasigroup AA of order mm, there is an isotope BB of Gm!+1G^{m!+1} such that

B∉V(A),B∉Q(A),B∉PV(A).B\notin\mathsf{V}(A),\qquad B\notin\mathsf{Q}(A),\qquad B\notin\mathsf{PV}(A).

Consequently, for C∈{V,Q,PV}\mathsf C\in\{\mathsf{V},\mathsf{Q},\mathsf{PV}\}, the class IC(G)I\mathsf C(G) is not generated under C\mathsf C by any single finite quasigroup.

A disjunctive identity is a universally quantified formula

⋁j=1rsj(x)=tj(x),(5) \tag{5} \bigvee_{j=1}^{r}s_j(\mathbf x)=t_j(\mathbf x),

where the sj,tjs_j,t_j are terms. Let D(A)\mathsf{D}(A) be the class of algebras satisfying every such formula valid in AA.

Theorem 2. Let C3C_3 be the cyclic group of order three. There is no finite quasigroup AA such that

D(A)=ID(C3).\mathsf{D}(A)=I\mathsf{D}(C_3).

Falconer (1, Theorem 5.2) constructs an isotope of a countable elementary abelian pp-group with an infinite one-generated subquasigroup. Her result already yields the variety and quasivariety obstructions in Theorem 1. We do not claim those conclusions as new. The proof below replaces the infinite construction by a cyclic coordinate shift on a finite direct power and applies equally to PV\mathsf{PV}. Theorem 2 uses a separate obstruction on quasigroups of order at most three.

A finite separating isotope

Define constant-free terms recursively by

t0(x,y)=x,tn+1(x,y)=tn(x,y)⋅y.(7) \tag{7} t_0(x,y)=x,\qquad t_{n+1}(x,y)=t_n(x,y)\cdot y.

If Ry(x)=x⋅yR_y(x)=x\cdot y, then tn(x,y)=Ryn(x)t_n(x,y)=R_y^n(x).

Lemma 3. Every quasigroup AA of finite order mm satisfies

tm!(x,y)=x.(8) \tag{8} t_{m!}(x,y)=x.

For a fixed nn, the identity tn(x,y)=xt_n(x,y)=x is preserved under homomorphic images, subalgebras and finite direct products.

Proof. The last two identities in (1) make RyR_y a permutation, with inverse z↦z/yz\mapsto z/y. Every permutation of an mm-element set has its m!m!-th power equal to the identity, which proves (8).

If f:A→Bf:A\to B is a homomorphism, induction on nn gives

f(tnA(x,y))=tnB(f(x),f(y)).f(t_n^A(x,y))=t_n^B(f(x),f(y)).

For surjective ff, lifting x,y∈Bx,y\in B proves preservation under images. For injective ff, cancellation proves preservation under subalgebras. In a direct product, the identity holds in each coordinate. ◻

Lemma 4. Let G≠1G\neq1 be a group and n≥1n\geq1. There is an isotope BnB_n of Gn+1G^{n+1} which does not satisfy tn(x,y)=xt_n(x,y)=x.

Proof. Put d=n+1d=n+1 and H=GdH=G^d, with coordinates indexed by Z/dZ\mathbb Z/d\mathbb Z. Let

(σx)i=xi+1.(\sigma x)_i=x_{i+1}.

On the set HH, define

x∗y=σ(x)y,x\∗z=σ(x)−1z,z/∗y=σ−1(zy−1).(11) \tag{11} x*y=\sigma(x)y,\qquad x\backslash_*z=\sigma(x)^{-1}z,\qquad z/_*y=\sigma^{-1}(zy^{-1}).

These operations satisfy (1), and (σ,1,1)(\sigma,1,1) is an isotopy from Bn=(H;∗,\∗,/∗)B_n=(H;*,\backslash_*,/_*) to HH.

Let ee be the identity of HH, choose g≠1g\neq1 in GG, and put

a0=g,ai=1(i≠0).a_0=g,\qquad a_i=1\quad(i\neq0).

Since x∗e=σ(x)x*e=\sigma(x), induction gives

tnBn(a,e)=σn(a).t_n^{B_n}(a,e)=\sigma^n(a).

Its zeroth coordinate is an=1a_n=1, whereas a0=g≠1a_0=g\neq1. Thus tnBn(a,e)≠at_n^{B_n}(a,e)\neq a. ◻

Proof of Theorem 1. Take n=m!n=m!. Lemma 3 gives an identity of AA, and hence of every member of V(A)\mathsf{V}(A) and Q(A)\mathsf{Q}(A). The preservation assertion gives the same identity throughout PV(A)\mathsf{PV}(A). The isotope BnB_n in Lemma 4 violates it.

Each of V(G)\mathsf{V}(G), Q(G)\mathsf{Q}(G) and PV(G)\mathsf{PV}(G) contains the finite direct power Gm!+1G^{m!+1}. Thus Bn∈IC(G)B_n\in I\mathsf C(G) but Bn∉C(A)B_n\notin\mathsf C(A) for all three operators. The argument applies to every finite proposed generator AA. ◻

Disjunctive identities

The argument in this section uses only a cardinality bound on the possible generator.

Lemma 5. Every nonempty quasigroup of order at most three satisfies at least one of the following identities:

(x⋅y)⋅z=x⋅(y⋅z),(x⋅y)⋅y=x,x⋅x=y⋅y.\begin{align*} (x\cdot y)\cdot z&=x\cdot(y\cdot z),\tag{14a}\\ (x\cdot y)\cdot y&=x,\tag{14b}\\ x\cdot x&=y\cdot y.\tag{14c} \end{align*}

Proof. For orders one and two the Latin square condition gives an associative operation. For order three, identify the underlying set with F3\mathbb F_3. The multiplication has the form

x⋅y=ax+by+c,a,b∈{1,−1},c∈F3.(15) \tag{15} x\cdot y=ax+by+c, \qquad a,b\in\{1,-1\},\quad c\in\mathbb F_3.

To see the exhaustion, permute rows and columns to make the first row and first column 0,1,20,1,2. The remaining entries are then forced, giving the table of addition modulo three. Every permutation of F3\mathbb F_3 is affine, so undoing the permutations gives (15).

If a=−1a=-1, then (x⋅y)⋅y=x(x\cdot y)\cdot y=x. If a=1a=1 and b=−1b=-1, then x⋅x=cx\cdot x=c. In the remaining case a=b=1a=b=1, the operation x+y+cx+y+c is associative. These are (14b), (14c), and (14a), respectively. ◻

Proof of Theorem 2. The disjunctive identity

⋁0≤i<j≤3xi=xj(16) \tag{16} \bigvee_{0\leq i<j\leq3}x_i=x_j

holds in C3C_3. Every group in D(C3)\mathsf{D}(C_3) therefore has at most three elements, and isotopy preserves cardinality.

Suppose D(A)=ID(C3)\mathsf{D}(A)=I\mathsf{D}(C_3). Since A∈D(A)A\in\mathsf{D}(A), it follows that ∣A∣≤3|A|\leq3. Moreover, C3∈D(C3)C_3\in\mathsf{D}(C_3), so the target class contains both addition on F3\mathbb F_3 and the isotope

x∘y=−x+y.x\circ y=-x+y.

Ordinary identities are special cases of disjunctive identities. Each identity in Lemma 5 is incompatible with containing both these quasigroups:

(0+1)+1=2≠0,0+0=0≠2=1+1,(1∘0)∘0=1≠−1=1∘(0∘0).\begin{align*} (0+1)+1&=2\neq0,\\ 0+0&=0\neq2=1+1,\\ (1\circ0)\circ0&=1\neq-1=1\circ(0\circ0). \end{align*}

Thus (14b) and (14c) exclude addition on F3\mathbb F_3, while (14a) excludes ∘\circ. Every possible AA is excluded. ◻

References

Preprint · Lean (GitHub)

  1. E. Falconer, Isotopy invariants in quasigroups, Trans. Amer. Math. Soc. 151 (1970), 511–526, doi:10.1090/S0002-9947-1970-0272932-4.
  1. A. A. Gvaramiya, Quasigroup classes that are invariant under isotopy, and abstract classes of invertible automata, Dokl. Akad. Nauk SSSR 282 (1985), no. 5, 1047–1051 (Russian), https://www.mathnet.ru/eng/dan9110.
  1. E. I. Khukhro and V. D. Mazurov (eds.), The Kourovka Notebook: Unsolved problems in group theory, 21st ed., September 2026 update, Problem 9.4, arXiv:1401.0300v46.