Finite traces of group varieties via Fn/⋂ker⁡fF_n/\bigcap\ker f and residual quotient closure

4.17

Problem

Characterize the classes K\mathcal K of finite groups that occur as all finite members of an ordinary group variety:

K=Vfin={Q∈V:∣Q∣<∞}.\mathcal K=\mathcal V_{\mathrm{fin}} =\{Q\in\mathcal V:|Q|<\infty\}.

The criterion below uses finite quotients of finitely generated residually-K\mathcal K groups. Residual separation allows arbitrary homomorphisms into members of K\mathcal K.

The residual criterion

An ordinary group variety is a class defined by word identities in the group signature. For such a variety V\mathcal V, write

Vfin={Q∈V:∣Q∣<∞}.\mathcal V_{\mathrm{fin}}=\{Q\in\mathcal V:|Q|<\infty\}.

Problem 4.17 of the Kourovka Notebook, due to R. Baer, asks for a characterization of the classes Vfin\mathcal V_{\mathrm{fin}} (2).

Let K\mathcal K be a class of finite groups. A group GG is residually-K\mathcal K if

∀g∈G∖{1}∃Q∈K∃f:G⟶Qf(g)≠1.(2) \tag{2} \forall g\in G\setminus\{1\}\quad \exists Q\in\mathcal K\quad\exists f:G\longrightarrow Q \quad f(g)\ne1.

All maps between groups in this article are homomorphisms. The maps in (2) need not be surjective. Consider the condition

G is finitely generated and residually-K,p:G↠Q,∣Q∣<∞}⟹Q∈K.(R) \tag{R} \left. \begin{gathered} G\text{ is finitely generated and residually-}\mathcal K,\\ p:G\twoheadrightarrow Q,\qquad |Q|<\infty \end{gathered} \right\}\quad\Longrightarrow\quad Q\in\mathcal K.

Theorem 1. For a class K\mathcal K of finite groups, the following are equivalent:

  1. K=Vfin\mathcal K=\mathcal V_{\mathrm{fin}} for an ordinary group variety V\mathcal V;

  2. K\mathcal K satisfies (R).

When these conditions hold, one may take V\mathcal V to be the class of all groups satisfying every word identity valid in K\mathcal K.

Here finite generation means the existence of an epimorphism Fn↠GF_n\twoheadrightarrow G, where Fn=F(x1,…,xn)F_n=F(x_1,\ldots,x_n) and n≥0n\ge0. In particular, the trivial group is allowed. No closure hypothesis on K\mathcal K is imposed. In the implication (1)⇒\Rightarrow(2), finite generation of GG is unnecessary.

Common laws and residual separation

Regard an identity as a pair (n,w)(n,w) with w∈Fnw\in F_n; its assertion in GG is

G⊨(n,w)⟺f(w)=1for every f:Fn⟶G.(4) \tag{4} G\models(n,w)\quad\Longleftrightarrow\quad f(w)=1\quad\text{for every }f:F_n\longrightarrow G.

This includes all ordinary group equations: the equation u=vu=v is written as the identity uv−1=1uv^{-1}=1. For a set Σ\Sigma of such pairs, let Mod⁡(Σ)\operatorname{Mod}(\Sigma) denote the groups satisfying every member of Σ\Sigma.

Lemma 2. Let Σ\Sigma be a set of ordinary identities.

  1. If every member of K\mathcal K satisfies Σ\Sigma, then every residually-K\mathcal K group satisfies Σ\Sigma.

  2. If GG satisfies Σ\Sigma and p:G↠Hp:G\twoheadrightarrow H, then HH satisfies Σ\Sigma.

Proof. For the first assertion, fix (n,w)∈Σ(n,w)\in\Sigma and f:Fn→Gf:F_n\to G. If f(w)≠1f(w)\ne1, residual separation gives Q∈KQ\in\mathcal K and q:G→Qq:G\to Q with qf(w)≠1qf(w)\ne1. This contradicts Q⊨(n,w)Q\models(n,w).

For the second assertion, let f:Fn→Hf:F_n\to H. Choose gi∈Gg_i\in G with p(gi)=f(xi)p(g_i)=f(x_i), and let f~:Fn→G\widetilde f:F_n\to G send xix_i to gig_i. The universal property of FnF_n gives pf~=fp\widetilde f=f. Consequently,

f(w)=p(f~(w))=p(1)=1((n,w)∈Σ).f(w)=p\bigl(\widetilde f(w)\bigr)=p(1)=1 \qquad ((n,w)\in\Sigma).

 ◻

For n≥0n\ge0, define

Rn(K)=⋂Q∈Kf:Fn→Qker⁡f,Ln(K)=Fn/Rn(K).(6) \tag{6} R_n(\mathcal K)= \bigcap_{\substack{Q\in\mathcal K\\ f:F_n\to Q}}\ker f, \qquad L_n(\mathcal K)=F_n/R_n(\mathcal K).

The intersection is a normal subgroup of FnF_n; an empty intersection means FnF_n. Put

ΣK={(n,w):n≥0, w∈Rn(K)}.(7) \tag{7} \Sigma_{\mathcal K} =\{(n,w):n\ge0,\ w\in R_n(\mathcal K)\}.

Thus ΣK\Sigma_{\mathcal K} is exactly the set of identities common to K\mathcal K. In particular,

Q∈K⟹Q⊨ΣK.(8) \tag{8} Q\in\mathcal K\quad\Longrightarrow\quad Q\models\Sigma_{\mathcal K}.

Lemma 3. The group Ln(K)L_n(\mathcal K) is finitely generated and residually-K\mathcal K.

Proof. The images of x1,…,xnx_1,\ldots,x_n generate Ln(K)L_n(\mathcal K). Let wRn(K)≠1wR_n(\mathcal K)\ne1. By (6), there are Q∈KQ\in\mathcal K and f:Fn→Qf:F_n\to Q such that f(w)≠1f(w)\ne1. Since Rn(K)≤ker⁡fR_n(\mathcal K)\le\ker f, the map ff induces

f‾:Ln(K)⟶Q,f‾(wRn(K))=f(w)≠1.\overline f:L_n(\mathcal K)\longrightarrow Q, \qquad \overline f\bigl(wR_n(\mathcal K)\bigr)=f(w)\ne1.

 ◻

Lemma 4. If G⊨ΣKG\models\Sigma_{\mathcal K} and p:Fn↠Gp:F_n\twoheadrightarrow G, then pp factors through an epimorphism Ln(K)↠GL_n(\mathcal K)\twoheadrightarrow G.

Proof. For w∈Rn(K)w\in R_n(\mathcal K), the pair (n,w)(n,w) belongs to ΣK\Sigma_{\mathcal K}, so (4) gives p(w)=1p(w)=1. Hence Rn(K)≤ker⁡pR_n(\mathcal K)\le\ker p, and the quotient universal property yields

p‾:Ln(K)⟶G,p‾(wRn(K))=p(w).\overline p:L_n(\mathcal K)\longrightarrow G, \qquad \overline p\bigl(wR_n(\mathcal K)\bigr)=p(w).

Every g∈Gg\in G is p(w)p(w) for some w∈Fnw\in F_n, proving surjectivity. ◻

Proof of the characterization

Proof of Theorem 1. Suppose K=Mod⁡(Σ)fin\mathcal K=\operatorname{Mod}(\Sigma)_{\mathrm{fin}}. Let GG be residually-K\mathcal K and let p:G↠Qp:G\twoheadrightarrow Q, where QQ is finite. By Lemma 2(1), G⊨ΣG\models\Sigma; by Lemma 2(2), Q⊨ΣQ\models\Sigma. Therefore Q∈KQ\in\mathcal K, proving (R).

Conversely, suppose (R) holds and set V=Mod⁡(ΣK)\mathcal V=\operatorname{Mod}(\Sigma_{\mathcal K}). Inclusion K⊆Vfin\mathcal K\subseteq\mathcal V_{\mathrm{fin}} is (8). If Q∈VfinQ\in\mathcal V_{\mathrm{fin}}, choose a finite generating tuple and its associated epimorphism Fn↠QF_n\twoheadrightarrow Q. Lemma 4 produces

Ln(K)↠Q.L_n(\mathcal K)\twoheadrightarrow Q.

Lemma 3 shows that its domain is finitely generated and residually-K\mathcal K. Condition (R) now gives Q∈KQ\in\mathcal K. Thus

K=Mod⁡(ΣK)fin,\mathcal K=\operatorname{Mod}(\Sigma_{\mathcal K})_{\mathrm{fin}},

as asserted. ◻

The quotient construction (6) is the group instance of the free-algebra construction in Birkhoff’s variety theorem (1).

References

Preprint · Lean (GitHub)

  1. G. Birkhoff, On the structure of abstract algebras, Proc. Cambridge Philos. Soc. 31 (1935), 433–454. doi:10.1017/S0305004100013463.
  1. E. I. Khukhro and V. D. Mazurov (eds.), Unsolved Problems in Group Theory: The Kourovka Notebook, 21st ed., 2026, Problem 4.17. arXiv:1401.0300.