Pure Graph Houghton Group

• Pure graph Houghton group is the subgroup of asymptotically rigid mapping classes that fixes every end individually in an infinite, locally finite graph.
• It is structured by an exact sequence extending the compactly supported mapping class group by a free abelian flux group of rank r-1, and is a finite-index subgroup of the full graph Houghton group.
• For r≥3, it is finitely presented with a solvable word problem and shares BNSR invariant characteristics with classical and surface Houghton groups.

The pure graph Houghton group is the end-fixing subgroup of a graph Houghton group, defined in the setting of asymptotically rigid mapping class groups of certain infinite graphs. For a connected, locally finite, infinite graph $\Gamma_r$ with $r$ ends, all accumulated by loops, the graph Houghton group $B(g,h,r)$ consists of asymptotically rigid mapping classes relative to a $(g,h)$-rigid structure, and the pure graph Houghton group is

$$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$

In the principal finitely ended case studied in detail, one writes

$PB_r:=PB(0,1,r).$

Thus $PB_r$ is the subgroup of asymptotically rigid mapping classes that fix each end individually. It is a finite-index subgroup of the full graph Houghton group $B_r=B(0,1,r)$, with

$[\,B_r:PB_r\,]=r!,$

and its basic algebraic structure is governed by the exact sequence

$1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$

For $r$0, $r$1 is finitely presented; more generally, it is of type $r$2 but not $r$3, is $r$4-ended for $r$5, has explicitly computable BNSR invariants, and for $r$6 has solvable word problem (Hill et al., 28 Aug 2025).

1. Ambient graph-theoretic framework

The ambient objects are connected, locally finite, infinite graphs $r$7, with mapping class group defined in the proper-homotopy category by

$r$8

and pure mapping class group $r$9 defined as the subgroup fixing the end space pointwise. The subgroup of compactly supported mapping classes is denoted

$B(g,h,r)$0

The end space is

$B(g,h,r)$1

for a compact exhaustion $B(g,h,r)$2. An end is accumulated by loops if every neighborhood of it has infinite rank (Hill et al., 28 Aug 2025).

For the finitely many ends theory, one fixes a graph $B(g,h,r)$3 with $B(g,h,r)$4 ends, all accumulated by loops, together with a rigid decomposition

$B(g,h,r)$5

Here $B(g,h,r)$6 is a finite-rank connected core of rank $B(g,h,r)$7, containing a center vertex $B(g,h,r)$8, with $B(g,h,r)$9 valence-$(g,h)$0 boundary vertices, and each $(g,h)$1 is a piece isomorphic to a model graph $(g,h)$2, where $(g,h)$3 is a finite graph of rank $(g,h)$4 obtained from a line segment with $(g,h)$5 vertices whose middle $(g,h)$6 vertices each carry one loop. A $(g,h)$7-rigid structure is such a decomposition with disjoint interiors and with each piece either disjoint from the core or attached along one boundary point $(g,h)$8. A suited subgraph is a connected union of the core with finitely many pieces (Hill et al., 28 Aug 2025).

A proper homotopy class represented by $(g,h)$9 is asymptotically rigid if there exists a suited subgraph $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$0 such that $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$1 is a proper homotopy equivalence onto its image, $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$2 is also a suited subgraph, and for every piece $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$3 there is a piece $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$4 with

$$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$5

Such a $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$6 is called a defining graph for $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$7. The graph Houghton group $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$8 is the group of asymptotically rigid mapping classes, and the pure graph Houghton group is the kernel of the action on the finite end space: $$PB(g,h,r):=\ker\big(B(g,h,r)\curvearrowright E(\Gamma_r)\big).$$9 In the standard $PB_r:=PB(0,1,r).$0-case this becomes $PB_r:=PB(0,1,r).$1 (Hill et al., 28 Aug 2025).

The term “pure” is therefore end-theoretic rather than graph-combinatorial: $PB_r:=PB(0,1,r).$2 consists exactly of asymptotically rigid mapping classes that fix each end individually. This parallels the pure surface Houghton group

$PB_r:=PB(0,1,r).$3

which fixes each end of the surface $PB_r:=PB(0,1,r).$4 and serves as the natural pure subgroup in the surface setting (Torgerson et al., 2024).

2. Structural description and the flux exact sequence

The central algebraic description of the pure graph Houghton group is the flux exact sequence. For $PB_r:=PB(0,1,r).$5, the flux map restricts to

$PB_r:=PB(0,1,r).$6

In the principal case $PB_r:=PB(0,1,r).$7, this becomes

$PB_r:=PB(0,1,r).$8

Thus $PB_r:=PB(0,1,r).$9 is an extension of the compactly supported mapping class group by a free abelian “flux” group of rank $PB_r$0. The loop shifts $PB_r$1 project to the standard basis of $PB_r$2 (Hill et al., 28 Aug 2025).

The coordinate flux maps are defined using coranks of free factors: $PB_r$3 This realizes the asymptotic part of a pure graph Houghton element as net loop flow between ends. A plausible implication is that the free abelian quotient plays the same structural role as the translation lattice in classical Houghton theory, where

$PB_r$4

for the $PB_r$5-ray set $PB_r$6 (Cox et al., 11 Aug 2025).

The compactly supported subgroup is described as a direct limit of finite-rank automorphism groups: $PB_r$7 through the exhaustion by finite subgraphs and Armstrong–Forrest–Vogtmann presentations. In particular, $PB_r$8 contains $PB_r$9 for every $B_r=B(0,1,r)$0. The subgroup generated by loop shifts

$B_r=B(0,1,r)$1

is isomorphic to the classical Houghton group $B_r=B(0,1,r)$2, so $B_r=B(0,1,r)$3 contains a canonical classical Houghton subgroup while also containing compactly supported automorphism-type subgroups modeled on $B_r=B(0,1,r)$4 (Hill et al., 28 Aug 2025).

Because $B_r=B(0,1,r)$5 has finitely many ends, $B_r=B(0,1,r)$6 is a finite-index subgroup of $B_r=B(0,1,r)$7. This finite-index relation mirrors the surface case, where

$B_r=B(0,1,r)$8

and

$B_r=B(0,1,r)$9

show the same basic pattern: a compactly supported kernel and a rank-$[\,B_r:PB_r\,]=r!,$0 asymptotic quotient (Aramayona et al., 2023).

3. Generators and explicit finite presentation

For $[\,B_r:PB_r\,]=r!,$1, the pure graph Houghton group is finitely generated. One explicit generating statement is

$[\,B_r:PB_r\,]=r!,$2

where the $[\,B_r:PB_r\,]=r!,$3 are loop shifts from $[\,B_r:PB_r\,]=r!,$4 to $[\,B_r:PB_r\,]=r!,$5, $[\,B_r:PB_r\,]=r!,$6 is a loop flip, and $[\,B_r:PB_r\,]=r!,$7 is a Nielsen-type automorphism supported near the first two loops. The full group satisfies

$[\,B_r:PB_r\,]=r!,$8

The proof proceeds by correcting an arbitrary pure element by suitable loop shifts until the resulting mapping class has zero flux and therefore lies in $[\,B_r:PB_r\,]=r!,$9; compactly supported generators are then recovered from shifted copies of $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$0, with $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$1 expressible as a commutator of loop shifts (Hill et al., 28 Aug 2025).

The main presentation theorem is that for $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$2,

$1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$3

where $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$4 is a finite set of relations indexed in the paper as $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$5. This gives an explicit finite presentation of $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$6, and hence shows that $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$7 is finitely presented for $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$8 (Hill et al., 28 Aug 2025).

The named generators have distinct geometric and algebraic roles. The $1\longrightarrow \Map_c(\Gamma_r)\longrightarrow PB_r\longrightarrow \mathbb Z^{r-1}\longrightarrow 1.$9 encode the asymptotic $r$00-part. The element $r$01 is a swap of two adjacent loops near the base end and satisfies the key identity

$r$02

The element $r$03 is a loop flip. The element $r$04 is a Nielsen-type automorphism; the paper records, for instance,

$r$05

fixing the other generators. The philosophy of the presentation is therefore a semidirect interaction between asymptotic loop shifts and compactly supported automorphism generators (Hill et al., 28 Aug 2025).

The construction of the presentation passes through an infinite presentation

$r$06

together with

$r$07

and then collapses the infinite AFV generator family to finitely many generators by rewriting the $r$08 as conjugates of $r$09 by loop shifts. A plausible implication is that the finite presentation is best viewed as a presentation of a compactly supported $r$10-type direct-limit part coupled to an asymptotic Houghton-type translation part.

4. Finiteness properties and BNSR invariants

The finiteness theory of pure graph Houghton groups is controlled by the number of ends. For graphs with $r$11 ends, all accumulated by loops, the graph Houghton group is of type

$r$12

Since $r$13 has finite index in $r$14, the same holds for the pure group: $r$15 Consequently, $r$16 is finitely generated for $r$17 and finitely presented for $r$18 (Hill et al., 28 Aug 2025).

The proof uses an action on a Stein–Farley-type cube complex $r$19. Its vertices are pairs $r$20, where $r$21 is a suited subgraph and $r$22, with Morse height given by the rank of $r$23. The complex is contractible, sublevel quotients are finite, and cube stabilizers are finite-index subgroups of finite-type graph mapping class groups and hence of type $r$24. Brown’s criterion then reduces the finiteness problem to the connectivity of descending links, which are analyzed via a piece complex and complete-join arguments (Hill et al., 28 Aug 2025).

The BNSR invariants of the pure graph Houghton group are computed in the same form as for several other Houghton-type groups. If $r$25 denotes either the pure graph Houghton group or the pure doubled handlebody Houghton group, and if a nonzero character is written in ascending standard form

$r$26

then

$r$27

Thus the BNSR layer of $r$28 is determined exactly by the number $r$29 of nonzero coefficients in ascending standard form (Hill et al., 28 Aug 2025).

This places the pure graph Houghton group in direct analogy with both classical and surface Houghton theory. For the classical Houghton groups $r$30, the complete BNSR computation has the same schematic form,

$r$31

with $r$32 defined from the ordered character coefficients (Zaremsky, 2018). The pure surface Houghton group $r$33 satisfies the corresponding statement

$r$34

proved via a CAT(0) Stein–Farley cube complex and an adaptation of Zaremsky’s method (Torgerson et al., 2024). The graph case therefore preserves the same BNSR combinatorics while replacing surface pieces by graph pieces.

5. Algebraic and large-scale properties

Several further structural properties of $r$35 are established. For $r$36, $r$37 is $r$38-ended. The proof in the full group case $r$39 explicitly also proves the same property for the pure subgroup. For $r$40, $r$41 has solvable word problem, proved using the explicit presentation together with reduction into the compactly supported subgroup and solvability of the word problem in $r$42 (Hill et al., 28 Aug 2025).

The group also exhibits a hybrid Houghton–automorphism profile. It contains $r$43 for every $r$44, yet also contains the classical Houghton subgroup generated by loop shifts. This suggests that $r$45 interpolates between two well-developed directions in geometric group theory: asymptotic end-translation groups and compactly supported automorphism groups of free groups. A plausible implication is that many of its geometric features arise from the interaction rather than from either factor in isolation.

The identity

$r$46

in the case $r$47 is emblematic. It shows that compactly supported loop-swapping behavior can be produced from asymptotic shifts alone, and it underlies both the finite-generation argument and the finite presentation (Hill et al., 28 Aug 2025).

The paper also records that graph Houghton groups contain free abelian subgroups of arbitrarily large rank. It further uses torsion phenomena in graph Houghton groups when comparing them with other Houghton-type families. For the pure subgroup, the strongest explicit statements are the finite presentation, one-endedness, BNSR computation, solvable word problem, and the embedding of all $r$48 (Hill et al., 28 Aug 2025).

6. Relation to other Houghton-type groups and terminology

The pure graph Houghton group belongs to a broader family of Houghton-type constructions, but it is not a rephrasing of earlier classical or surface notions. The paper introducing $r$49 states that graph Houghton groups are not commensurable with the classical, surface, braided, handlebody, and doubled handlebody Houghton groups, and uses pure subgroups in the proofs of these separations. At the same time, the formal headline theorem is stated for the full groups $r$50, not as a standalone theorem asserting pairwise non-commensurability of all pure variants (Hill et al., 28 Aug 2025).

A common misconception is to identify $r$51 with a “pure surface Houghton group.” The pure surface Houghton group is instead

$r$52

the subgroup of asymptotically rigid mapping classes of an infinite-type surface $r$53 that fix each end. It has abelianization

$r$54

and its BNSR invariants were computed independently (Torgerson et al., 2024). Similarly, the groups

$r$55

from the surface classification theory are pure surface Houghton groups, classified up to isomorphism by $r$56 and up to commensurability by $r$57 (Aramayona et al., 2023). These are surface analogues, not graph Houghton groups.

Another potential source of confusion is the use of graph-theoretic techniques in ordinary Houghton-group papers. For example, the 2025 paper on full-Hirsch-length subgroups of classical Houghton groups analyzes graph-wreath products and restricted multi-wreath products, but explicitly does not define or study an object called a pure graph Houghton group; there the graph input is auxiliary to finiteness arguments about subgroups of classical $r$58 (Cox et al., 11 Aug 2025). Likewise, earlier papers on classical Houghton groups, hyperbolic structures, and Houghton-like groups from shift-similar groups develop ray-based or cube-complex generalizations, but not the end-fixing asymptotically rigid graph-mapping-class groups denoted $r$59 (Genevois et al., 18 Feb 2025).

The most precise way to situate the pure graph Houghton group is therefore as follows. It is the end-pure subgroup of an asymptotically rigid mapping class group of an infinite graph with finitely many ends, all accumulated by loops. It admits the exact sequence

$r$60

has an explicit finite presentation for $r$61, is of type $r$62 but not $r$63, and shares the same BNSR character formula as classical and pure surface Houghton groups. At the same time, the surrounding graph Houghton theory defines a genuinely new commensurability class among Houghton-type groups, so $r$64 is best understood as a new graph-based pure Houghton construction rather than as a reformulation of an existing one (Hill et al., 28 Aug 2025).