---
title: Coset Identity in Group Theory
url: https://www.emergentmind.com/topics/coset-identity
type: topic
---

# Coset Identity in Group Theory

Searching arXiv for the primary paper and related usages of “coset identity”.
In group theory, a coset identity is a nontrivial word that vanishes identically on a product of cosets of a finite-index subgroup, or, in the topological setting, of an open subgroup. In the framework of $\sigma$-compact $K$-analytic groups over a non-archimedean local field, the notion is tightly linked to probabilistic identities: positive Haar measure of the identity fiber of a word map forces the existence of such a coset-level vanishing set [2507.19086]. The term also appears with distinct technical meanings in several adjacent areas, but the group-theoretic usage is the one that underlies recent results on pro-$p$ groups, compact linear groups over local fields, and probabilistic Tits alternatives [2507.19086].

## 1. Definition in topological and profinite group theory

Let $F_m$ be the free abstract group of rank $m \ge 1$, and let $w=w(x_1,\dots,x_m)\in F_m$ be a nontrivial group word. For a topological group $G$, the associated word map is
$$
w_G:G^m\to G,\qquad (g_1,\dots,g_m)\mapsto w(g_1,\dots,g_m).
$$
When $G$ is a $K$-analytic group, this map is analytic. A word $w$ is a coset identity in $G$ if there exist a finite-index subgroup $H\le G$ and elements $g_1,\dots,g_m\in G$ such that
$$
\forall h_1,\dots,h_m\in H,\qquad w(h_1g_1,\dots,h_mg_m)=1_G.
$$
In the topological variant used for profinite and analytic groups, one replaces finite-index by open; then $w$ is an open coset identity if
$$
\exists\, H\le_o G,\ \exists\, g_1,\dots,g_m\in G,\ \forall h_1,\dots,h_m\in H,\qquad
w(h_1g_1,\dots,h_mg_m)=1_G.
$$
For profinite groups, “coset identity” and “open coset identity” are equivalent [2507.19086].

The corresponding measure-theoretic notion is a probabilistic identity. If $G$ is a $\sigma$-compact $K$-analytic group with Haar measure $\mu$, then $w$ is a probabilistic identity if
$$
\mu^m\bigl(w_G^{-1}(1_G)\bigr)>0.
$$
For profinite $G$, the notation $P(G,w):=\mu_{G^m}(w^{-1}(1))$ is used, and $w$ is a probabilistic identity when $P(G,w)>0$. Since $w_G$ is continuous, the fiber $w_G^{-1}(1_G)$ is closed and therefore measurable [2507.19086].

## 2. Analytic rigidity: probabilistic identity implies coset identity

The central analytic result is Theorem A: if $G$ is a $\sigma$-compact $K$-analytic group and $w\in F_m$ is nontrivial, then
$$
\mu^m\bigl(w^{-1}(1_G)\bigr)>0
\quad\Longrightarrow\quad
\exists\, H\le_o G,\ \exists\, g_1,\dots,g_m\in G,\ \forall h_1,\dots,h_m\in H,\ 
w(h_1g_1,\dots,h_mg_m)=1_G.
$$
The statement also applies to words with parameters, because they define analytic maps as well [2507.19086].

The proof rests on an analytic-geometric dichotomy for fibers of analytic maps. For an analytic map $f:M\to N$ between $K$-analytic manifolds and a point $y\in N$, the fiber $f^{-1}(y)$ is either locally negligible, meaning measure zero near every point, or has non-empty interior. The argument uses local analytic charts, reduction to convergent power series, reduction to a single coordinate function, the Weierstrass Preparation Theorem, and Fubini’s theorem. The outcome is that positive measure of a fiber forces interior [2507.19086].

Applied to the word map $f=w_G$ and $y=1_G$, this yields that $w^{-1}(1_G)$ contains a product of cosets of an open subgroup,
$$
(g_1N)\times\cdots\times(g_mN)\subseteq w^{-1}(1_G),
$$
which is exactly an open coset identity with $H=N$. In profinite groups, where open subgroups are precisely the finite-index subgroups, this recovers the equivalence between probabilistic identity and coset identity [2507.19086].

## 3. Structural consequences and probabilistic Tits alternatives

One consequence is a probabilistic Tits alternative for compact linear groups over local fields. If $K$ is a local field, archimedean or non-archimedean, and $G$ is a compact subgroup of $\mathrm{GL}_n(K)$, then $G$ is either virtually solvable or randomly free. In particular, a $p$-adic analytic pro-$p$ group is either solvable or randomly free [2507.19086]. The mechanism is that any probabilistic identity is, by Theorem A, a coset identity, while the existence of a dense free subgroup in the non-virtually solvable case excludes coset identities.

A parallel statement holds for finitely generated linear groups. If $\Gamma$ is a finitely generated linear group, then $\Gamma$ is either virtually solvable or randomly free; moreover, any probabilistic identity, possibly with parameters, is a coset identity. The argument passes through an embedding into $\mathrm{GL}_n(A)$ for a finitely generated integral domain $A$, then into the ring of integers of a local field, and finally to the compact analytic closure [2507.19086].

The same dichotomy extends to several pro-$p$ classes obtained from free constructions. If $G$ is a virtually free pro-$p$ group, then $G$ is either virtually procyclic or randomly free. For finitely generated Demushkin groups, nontrivial free pro-$p$ products, and pro-$p$ analogues of limit groups in the class $\mathcal L$, one likewise has: either $G$ is solvable, or $G$ is randomly free. The proofs use surjections onto virtually free non-abelian pro-$p$ groups, induction or representation-theoretic arguments, and the fact that random freeness can be pushed back along quotients [2507.19086].

## 4. Torsion probabilistic identities and Lie-theoretic characterization

A torsion probabilistic identity is a probabilistic identity for a power word $w(x)=x^n$; equivalently, on a compact group $G$,
$$
\mu\bigl(\{x\in G:x^n=1_G\}\bigr)>0.
$$
For compact $p$-adic analytic groups, the paper gives a Lie-theoretic characterization. The following are equivalent:

$$
\text{(i)}\ \mu(\operatorname{Tor}(G))>0,
\qquad
\text{(ii)}\ \exists\, t\in\operatorname{Tor}(G)\ \text{such that } c_t \text{ is uniformly fixed-point-free on } G,
$$
$$
\text{(iii)}\ \exists\, t\in\operatorname{Tor}(G)\ \text{such that } dc_t \text{ is fixed-point-free on } \mathcal L(G).
$$

Here $c_t$ is conjugation by $t$, and $\mathcal L(G)$ is the Lie algebra of $G$. The implication $(i)\Rightarrow(ii)$ comes from Theorem A: a torsion probabilistic identity yields an open coset $tH$ consisting of torsion, and for torsion-free uniform $H$ this forces $C_H(t)=1$. The equivalence $(ii)\Leftrightarrow(iii)$ transfers fixed-point-free behavior between the group and the Lie algebra. For $(ii)\Rightarrow(i)$, one studies
$$
f(g)=g^tg^{-1},
$$
whose differential is $df=dc_t-\mathrm{id}$; invertibility makes $f$ locally surjective near $1_G$, so an open subgroup lies in $f(G)$, and the corresponding coset consists of torsion [2507.19086].

Several consequences follow. If $G$ is a compact $p$-adic analytic group that is not virtually solvable, then $\mu(\operatorname{Tor}(G))=0$. More specifically, when $\mu(\operatorname{Tor}(G))>0$ for a $p$-adic analytic pro-$p$ group, the group is solvable and admits an open uniform subgroup $U$ and a torsion element $t$ with $C_U(t)=1$. If $x^p$ is a probabilistic identity, then $\mathcal L(G)$ is nilpotent, hence $U$ is nilpotent [2507.19086].

For countably based profinite groups, torsion conjugacy classes can also be measured explicitly. If $G$ has a normal finitely generated non-abelian free pro-$p$ subgroup $F\le G$, then for any torsion $t\in G$ one has $\mu(t^G)=0$. More generally, for $x\in G$,
$$
\mu(x^G)=\frac{1}{c(x)},
\qquad
c(x)=\sup_{N\trianglelefteq_o G}\,|C_{G/N}(xN)|.
$$
The proof uses Lie ring methods, lower central series, and fixed-point arguments on nilpotent quotients of $F$ [2507.19086].

## 5. Examples, counterexamples, and hypotheses

The commutator word $w(x,y)=[x,y]$ gives a basic test case. If $G$ is a $\sigma$-compact $K$-analytic group and
$$
\mu^2\bigl(\{(x,y):[x,y]=1\}\bigr)>0,
$$
then there exist an open subgroup $H\le G$ and elements $g_1,g_2\in G$ such that
$$
\forall h_1,h_2\in H,\qquad [h_1g_1,h_2g_2]=1.
$$
Thus the cosets $g_1H$ and $g_2H$ commute pairwise. In many settings, if one of these cosets generates an open subgroup, this forces an open abelian subgroup. By contrast, in compact linear groups that are not virtually solvable, the set of commuting pairs has Haar measure zero [2507.19086].

Concrete examples illustrate the rigidity. $G=\mathrm{SL}_2(\mathbb Z_p)$ is a compact $p$-adic analytic group that is not virtually solvable, so it is randomly free. Therefore no nontrivial word, including $[x,y]$ or $x^n$, has positive-measure identity fiber; in particular, $\mu(\operatorname{Tor}(G))=0$ and
$$
\mu^2\bigl(\{(x,y):[x,y]=1\}\bigr)=0.
$$
Likewise, non-abelian free pro-$p$ groups of finite rank at least $2$ are randomly free unless virtually procyclic, so they admit no probabilistic identities and no positive-measure torsion sets [2507.19086].

The analytic theorem depends on regularity assumptions that are not formalities. The fiber dichotomy uses $\sigma$-compactness to pass from local negligibility to global measure zero via countable subcovers. Haar measure on $G$ and its product measure on $G^m$ are essential, and the argument uses local non-archimedean analytic geometry, power series in charts, Weierstrass Preparation, and analytic implicit-function arguments. In the profinite case, quotient maps decrease measure, so randomly free behavior can be lifted from quotients back to the original group [2507.19086].

## 6. Other technical meanings of the term

The expression “coset identity” is not uniform across the literature. In descriptive-topological group theory, the closely related problem is the existence of a common transversal for left and right coset spaces. If $G$ is Polish and $H\le G$ is compact, and if
$$
[H:H\cap xHx^{-1}]=[H:H\cap x^{-1}Hx]\qquad\text{for all }x\in G,
$$
then there exists a Borel set $T\subset G$ such that
$$
G=\bigsqcup_{t\in T}tH=\bigsqcup_{t\in T}Ht,
$$
so $T$ realizes a Borel bijection between $G/H$ and $H\backslash G$ [2305.02612]. This is a statement about simultaneous left-right coset representatives, not about word identities.

In $(2+1)$-dimensional topological order, gauging a non-normal subgroup $H\subset G$ produces a coset non-invertible symmetry whose identity object is the condensation defect
$$
\tilde U_{[1]}=C_{\mathrm{Rep}(H)}.
$$
Within the sandwich construction,
$$
\tilde U_{[g]}=D_{\mathrm{Rep}(H)}\times U_g\times \overline D_{\mathrm{Rep}(H)},
$$
and the fusion rule is
$$
\tilde U_{[g]}\times \tilde U_{[g']}=\sum_{h\in H}\tilde U_{[ghg'h^{-1}]},
$$
with $C_{\mathrm{Rep}(H)}$ as the tensor unit after canonical normalization [2405.20401].

In categorical coset constructions and vertex-operator-algebra theory, the phrase refers to field identification and selection rules. The Kac–Wakimoto set
$$
\mathrm{KW}:=\{W^\beta\in O(C_1)\mid A\boxtimes_{C_1\boxtimes C_2}(W^\beta\boxtimes 1_{C_2})\in C\}
$$
generates the identification group, while the selection rule is expressed by a Müger-centralizer condition. If $\mathrm{KW}$ is cyclic, multiplicities are $1$, and coset labels are identified along $\mathrm{KW}$-orbits [2404.00778].

In skew lattices, coset identities are flat coset laws rather than word identities. For comparable $D$-classes $A>B$, one has factorization formulas such as
$$
A\wedge x\wedge A \cong (A\wedge x)\times (x\wedge A),
$$
and, under the stated hypothesis on full cosets,
$$
(x\wedge A)\cap (A\wedge x')=\{x\wedge x'\}.
$$
These laws control normality, quasi-normality, and cancellation properties [1406.2510]. In coset relation algebras, the characteristic identity is a coset-shifted multiplication law,
$$
R_{xy,\alpha};R_{yz,\beta}
=
\bigcup\{R_{xz,\gamma}:H_{xz,\gamma}\subseteq \phi_{xy}[K_{xy,\alpha}\cdot H_{yz,\beta}]\cdot C_{xyz}\},
$$
which defines the class of full coset relation algebras and underlies the variety theorem for coset relation algebras [1804.03524]. Related but different uses also occur in coset-based constructions of supersymmetric Born–Infeld theory, where a covariant Bianchi identity is derived from Maurer–Cartan constraints, and in symmetric-space supergravity reductions, where the identity point is encoded by a basepoint matrix $M_0$ of a coset representative [1505.07386] [1301.3028].

Source: https://www.emergentmind.com/topics/coset-identity