Higman Embeddings in Group Theory

• Higman embeddings are results that characterize recursively presented groups via embeddings into finitely presented groups using techniques like HNN extensions.
• They provide explicit algorithmic constructions and refinements, preserving key geometric and subgroup properties such as quasi-isometric and malnormal embeddings.
• Recent advances extend the theory to Boone–Higman settings and specialized varieties, linking recursion theory, geometric group theory, and algorithmic presentations.

Higman embeddings are the family of embedding results that begin with Higman’s Embedding Theorem: a finitely generated group is recursively presented if and only if it embeds in a finitely presented group. In later work, the term extends naturally to refinements and descendants of that theorem, including explicit constructive embeddings, quasi-isometric and malnormal embeddings, Boone–Higman-type embeddings into finitely presented simple groups, and variants inside restricted classes such as Burnside varieties. A related but distinct strand studies embedding obstructions for Higman’s own group into metric ultraproducts of finite groups with controlled length functions (Belk et al., 2023, Mikaelian, 6 Jul 2025, Thom, 2010).

1. Classical theorem and the HNN mechanism

The classical statement is: every finitely generated recursively presented group embeds in a finitely presented group, and conversely every finitely generated subgroup of a finitely presented group is recursively presented (1908.10153). In the standard formulation used repeatedly in the recent literature, a finitely generated group $G$ is recursively presented when $G=\langle X\mid R\rangle$ with $X$ finite and $R\subseteq F(X)$ recursively enumerable; equivalently, there is a Turing machine enumerating the relators (Mikaelian, 6 Jul 2025).

The historical mechanism behind the theorem is the HNN extension. Given a group $G$, subgroups $A,B\leq G$, and an isomorphism $\phi:A\to B$, the HNN extension is

$$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$

with $t$ the stable letter (Bridson et al., 11 Dec 2025). Britton’s Lemma gives the normal-form control that makes the natural map $G\to G*_{\phi}$ injective, and this injectivity is one of the central engines of Higman-type embedding arguments (Bridson et al., 11 Dec 2025).

The 1949 Higman–Neumann–Neumann paper already supplied two foundational embedding results. One is that specified isomorphisms between subgroups can be realized by conjugation inside a larger group. Another is that any countable group embeds in a $G=\langle X\mid R\rangle$0-generator group, and if the original group has $G=\langle X\mid R\rangle$1 defining relations then the $G=\langle X\mid R\rangle$2-generator overgroup can be chosen with $G=\langle X\mid R\rangle$3 defining relations (Bridson et al., 11 Dec 2025). Higman’s 1961 theorem transformed these constructions into a characterization of recursively presented groups.

In modern expositions, the proof is usually described as an encoding of recursively enumerable relations by iterated HNN-extensions and free products with amalgamation, with Britton’s Lemma and normal forms preventing unwanted collapses (Belk et al., 2023). This suggests that “Higman embedding” is not a single construction but a general paradigm in which recursive data are converted into finitely many generators and relations by controlled free constructions.

2. Benign subgroups, Higman operations, and the rope trick

A central reformulation of Higman’s theorem is in terms of benign subgroups. In the formulation used by the recent modified proof, a subgroup $G=\langle X\mid R\rangle$4 of a finitely generated group $G=\langle X\mid R\rangle$5 is benign in $G=\langle X\mid R\rangle$6 if $G=\langle X\mid R\rangle$7 embeds in a finitely presented group $G=\langle X\mid R\rangle$8 having a finitely generated subgroup $G=\langle X\mid R\rangle$9 with $X$0 (1908.10153). This notion converts the problem of embedding recursively presented quotients into a subgroup-intersection problem inside finitely presented overgroups.

The sequence-theoretic side of the construction is built from the set $X$1 of all functions $X$2 with finite support. Higman introduced operations

$X$3

on subsets of $X$4, and a key theorem identifies recursively enumerable subsets of $X$5 with those generated from basic seeds by these operations (1908.10153). In the free group $X$6, one sets $X$7, defines $X$8, then $X$9, and finally $R\subseteq F(X)$0. The central equivalence is that $R\subseteq F(X)$1 is recursively enumerable if and only if $R\subseteq F(X)$2 is benign in $R\subseteq F(X)$3 (1908.10153).

The modified proof of Higman’s theorem shortens the original argument by replacing several homomorphism lemmas with combinatorial observations on words in HNN-extensions and free products with amalgamation (1908.10153). The later “auxiliary-free” star construction

$R\subseteq F(X)$4

packages nested HNN-extensions and amalgamated free products into a single finitely presented overgroup, and it yields general lemmas showing that finite intersections and joins of benign subgroups remain benign (Mikaelian, 2023). This is a direct generalization of a series of structures used by Higman for embeddings of recursive groups.

The final step is the “Higman rope trick.” Starting from a benign subgroup $R\subseteq F(X)$5, one builds a finitely presented overgroup in which $R\subseteq F(X)$6 embeds, and then uses an HNN-extension to eliminate infinitely many relations while preserving injectivity (Mikaelian, 6 Jul 2025). In this form, the theorem becomes a pipeline: recursively enumerable relations $R\subseteq F(X)$7 sequence sets $R\subseteq F(X)$8 benign subgroups $R\subseteq F(X)$9 finitely presented ambient group.

3. Explicit and algorithmic constructions

Recent work turns Higman’s existential theorem into an explicit algorithm. Given a recursive group $G$0, specified by effectively enumerable generators and recursively enumerable relations, one can output an explicit embedding of $G$1 into a finitely presented group, and the finitely presented target can even be chosen to be $G$2-generator (Mikaelian, 6 Jul 2025). The input is a recursive presentation $G$3; the output is a finitely presented group $G$4 together with an explicit embedding $G$5.

The construction begins with “universal words,” embedding $G$6 into a $G$7-generator recursive group $G$8. It then rewrites each relator $G$9 as an alternating product of powers of $A,B\leq G$0 and $A,B\leq G$1, thereby encoding it by an integer sequence $A,B\leq G$2 and forming a recursively enumerable set $A,B\leq G$3 (Mikaelian, 6 Jul 2025). From $A,B\leq G$4, the construction builds the subgroup

$A,B\leq G$5

and explicitly constructs finitely presented overgroups $A,B\leq G$6 and finitely generated subgroups $A,B\leq G$7 satisfying $A,B\leq G$8 (Mikaelian, 6 Jul 2025).

A notable feature of this algorithmic version is that it makes each Higman operation explicit. Free products with amalgamation, HNN-extensions, and the star construction are given as concrete steps after every operation in

$A,B\leq G$9

so the entire construction remains inside explicitly written finitely presented groups (Mikaelian, 6 Jul 2025). The final finitely presented target $\phi:A\to B$0 has $\phi:A\to B$1 generators and $\phi:A\to B$2 defining relations, where $\phi:A\to B$3 is the number of generators of $\phi:A\to B$4, $\phi:A\to B$5 is the number of defining relations of $\phi:A\to B$6, and $\phi:A\to B$7 is the number of generators of $\phi:A\to B$8 (Mikaelian, 6 Jul 2025).

The additive group of rationals is a canonical worked example. It has the recursive presentation

$\phi:A\to B$9

and the explicit algorithm constructs both a finitely presented group $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$0 with $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$1 and a $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$2-generator finitely presented group $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$3 with $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$4 (Mikaelian, 6 Jul 2025). Closely related work on Higman operations emphasizes that this explicit sequence machinery is particularly effective for free abelian, metabelian, soluble, nilpotent, divisible abelian, quasicyclic, and rational additive groups (Mikaelian, 2020, Mikaelian, 2023).

4. Geometric and subgroup-structural refinements

A major modern development is that Higman embeddings can be refined to preserve substantial geometric and subgroup-theoretic structure. One refinement proves that for any finitely generated recursively presented group $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$5, there exists a finitely presented group $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$6 and a malnormal embedding $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$7 (Wagner, 2024). Stronger versions show that $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$8 can be chosen to be a CEP-subgroup and that the restriction of the word-length on $$\langle G,t \mid t^{-1} a t=\phi(a)\text{ for all }a\in A\rangle,$$9 to $t$0 is equivalent to the word-length on $t$1; in the paper’s terminology, there exist constants $t$2 such that

$t$3

for all $t$4 (Wagner, 2024). If $t$5 has decidable Word Problem, then $t$6 can be chosen with decidable Word Problem as well, yielding a refinement of a theorem of Clapham (Wagner, 2024).

A second line of refinement links Higman embeddings to Dehn functions and the complexity of the Word Problem. If a finitely generated group has a presentation whose relators can be enumerated by a computational model satisfying certain technical requirements, then the group embeds quasi-isometrically into a finitely presented group whose Dehn function is bounded above by a function of the model’s computational complexity and the Dehn function of the original presentation (Wagner, 22 Sep 2025). In the main theorem of the paper’s introduction, if $t$7 for a multi-tape nondeterministic Turing machine enumerating the identity words and $t$8 is superadditive, then for any $t$9 there exists a quasi-isometric embedding into a finitely presented group $G\to G*_{\phi}$0 such that

$G\to G*_{\phi}$1

This improves the BORS bound and strengthens the embedding to a quasi-isometric one (Wagner, 22 Sep 2025).

These refinements shift the subject from pure existence to controlled existence. A plausible implication is that the modern theory of Higman embeddings now sits at the intersection of recursion theory, geometric group theory, and subgroup separability phenomena, rather than being only an existence theorem about finite presentations.

5. Boone–Higman descendants and simple targets

The Boone–Higman conjecture replaces “finitely presented” by “finitely presented simple.” In the formulation used by the modern survey literature, a finitely generated group $G\to G*_{\phi}$2 has solvable word problem if and only if it embeds as a subgroup of some finitely presented simple group (Belk et al., 2023). The easy direction is known: if $G\to G*_{\phi}$3 embeds into a finitely presented simple group, then $G\to G*_{\phi}$4 has solvable word problem (Belk et al., 2023).

A large part of current research studies Boone–Higman-type embeddings for specific classes. Every hyperbolic group embeds as a subgroup of a finitely presented simple group, proved by embedding hyperbolic groups into full, contracting rational similarity groups and then into finitely presented simple twisted Brin–Thompson groups (Belk et al., 2023). Every contracting self-similar group embeds into a finitely presented simple group through the chain

$G\to G*_{\phi}$5

where $G\to G*_{\phi}$6 is the Röver–Nekrashevych group and $G\to G*_{\phi}$7 is a twisted Brin–Thompson group (Belk et al., 2024).

The left-orderable setting gives a different refinement. Every countable left-ordered group embeds into a finitely generated left-ordered simple group, the embedding can be chosen to be a Frattini embedding, and if the source group is finitely generated the embedding is isometric (Darbinyan et al., 2020). Moreover, if the original order is computable, then the target group can be chosen to be computably left-ordered as well (Darbinyan et al., 2020).

Recent results also place $G\to G*_{\phi}$8 and several related families inside the Boone–Higman framework. For each $G\to G*_{\phi}$9, $G=\langle X\mid R\rangle$00 embeds in a finitely presented simple group, in fact in a twisted Brin–Thompson group $G=\langle X\mid R\rangle$01, so $G=\langle X\mid R\rangle$02 satisfies the Boone–Higman conjecture and its “permutational” variant (Belk et al., 27 Mar 2025). As consequences, the same holds for many groups that embed or virtually embed into some $G=\langle X\mid R\rangle$03, including mapping class groups of non-closed surfaces, braid groups, loop braid groups, ribbon braid groups, and certain Artin groups (Belk et al., 27 Mar 2025). The same paper proves that finitely presented twisted Brin–Thompson groups are universal among finitely presented simple highly transitive groups (Belk et al., 27 Mar 2025).

Taken together, these results show that Boone–Higman embeddings are no longer limited to a few classical containers. Hyperbolic groups, contracting self-similar groups, left-orderable groups, and automorphism groups of free groups now admit highly structured simple overgroups with additional control such as high transitivity, left-orderability, Frattini behavior, or isometry.

6. Variants, limits, and adjacent embedding theories

One specialized variant places Higman’s theorem inside the Burnside variety $G=\langle X\mid R\rangle$04. For all sufficiently large odd integers $G=\langle X\mid R\rangle$05, a finitely generated group $G=\langle X\mid R\rangle$06 from $G=\langle X\mid R\rangle$07 has a presentation $G=\langle X\mid R\rangle$08 with a finite set of generators $G=\langle X\mid R\rangle$09 and a recursively enumerable set $G=\langle X\mid R\rangle$10 of defining relations if and only if it is a subgroup of a group $G=\langle X\mid R\rangle$11 finitely presented in the variety $G=\langle X\mid R\rangle$12 (Olshanskii, 2019). The same work deduces a universal $G=\langle X\mid R\rangle$13-generated finitely presented in $G=\langle X\mid R\rangle$14 group and a $G=\langle X\mid R\rangle$15-generated finitely presented in $G=\langle X\mid R\rangle$16 group with undecidable word problem (Olshanskii, 2019).

A sharp limitation appears in metric approximation theory. Higman’s group

$G=\langle X\mid R\rangle$17

does not have the $G=\langle X\mid R\rangle$18-approximation property, equivalently it does not embed into any metric ultraproduct of finite groups equipped with commutator-contractive invariant length functions (Thom, 2010). The obstruction is quantitative: approximate Higman relations in such finite metric groups force the generator lengths to collapse below $G=\langle X\mid R\rangle$19 when $G=\langle X\mid R\rangle$20 (Thom, 2010). The result excludes only the restrictive class $G=\langle X\mid R\rangle$21, so it does not resolve whether Higman’s group is sofic or hyperlinear (Thom, 2010).

The term “Higman embeddings” also occurs in nearby but distinct settings involving Higman–Thompson groups. All Higman–Thompson groups $G=\langle X\mid R\rangle$22 embed into one another, extending the embeddings given by Higman in 1974 (Birget, 2019). In a different algebraic direction, the Higman–Thompson group of an unfolding tree $G=\langle X\mid R\rangle$23 embeds into the unitary group $G=\langle X\mid R\rangle$24 of the rooted Leavitt path algebra via

$G=\langle X\mid R\rangle$25

and over $G=\langle X\mid R\rangle$26 this yields that any $G=\langle X\mid R\rangle$27-isomorphism of rooted Leavitt path algebras induces an isomorphism of the associated Higman–Thompson groups (Gorazd, 2 Apr 2025).

These variants clarify the scope of the subject. In its strict classical sense, a Higman embedding is an embedding of a recursively presented group into a finitely presented group. In broader modern usage, it includes controlled embeddings into finitely presented simple groups, explicit algorithmic constructions, embeddings inside restricted varieties, and negative results that delimit which ambient approximation classes are unavailable.