Halin's Grid Theorem

• Halin's Grid Theorem is a structural result in infinite graph theory that shows a thick end forces a subdivision of a grid-like graph.
• It connects the existence of infinitely many disjoint rays within an end to the presence of hexagonal or square grid minors, impacting geometric structure.
• Recent refinements prescribe specific ray families and employ coarse geometric techniques to extend the classical theorem to broader settings.

Searching arXiv for recent and foundational papers on Halin's Grid Theorem and its coarse/generalized variants. arxiv_search.query({"search_query":"all:\"Halin's Grid Theorem\" OR ti:\"Halin\" grid theorem", "start":0, "max_results":10, "sort_by":"submittedDate", "sort_order":"descending"}) Halin's Grid Theorem is a classical structural result in infinite graph theory that links the geometry of an end to the existence of an infinite grid-like subgraph. In its modern formulations, the theorem states that a graph with a thick end—an end containing infinitely many pairwise disjoint rays—must contain a subdivision of a hexagonal half-grid or quarter grid whose vertical rays lie in that end. Subsequent work has turned this statement into a broader framework: one can prescribe the rays used as grid columns, upgrade half-grids to full hexagonal grids under quasi-transitivity, formulate quasi-isometry-invariant coarse analogues via asymptotic minors, characterize full square-grid minors by bundle structure, and extend the theorem to directed ends and to selected uncountable-degree settings (Kurkofka et al., 2021, Georgakopoulos et al., 2024, Albrechtsen et al., 17 Jul 2025, Reich, 2024).

1. Classical theorem and foundational notions

A ray is a one-way infinite simple path, and a tail of a ray is any subray obtained by deleting a finite initial segment. Two rays are equivalent if they cannot be separated by a finite vertex set; equivalently, in the formulations used in the recent literature, if there exist infinitely many pairwise vertex-disjoint paths between them. An end is an equivalence class of rays. The degree of an end is the maximum cardinality of a family of pairwise disjoint rays contained in that end, and an end is thick if this degree is infinite (Kurkofka et al., 2021, Geschke et al., 2020).

In a standard modern form, Halin's Grid Theorem asserts that every graph with a thick end contains a subdivision of the hexagonal half-grid whose vertical rays belong to that end. A recent short-proof formulation states equivalently that an infinite graph has a thick end if and only if the hexagonal quarter grid $H^\infty$ is a topological minor of the graph (Chern, 12 Sep 2025). This theorem is independent of planarity assumptions on the ambient graph, and modern expositions emphasize that no local finiteness assumption is needed for the topological-minor version (Kurkofka et al., 2021, Chern, 12 Sep 2025).

The theorem sits on top of a compactness phenomenon already proved by Halin: if a graph contains $n$ pairwise disjoint rays for every $n \in \mathbb N$, then it contains infinitely many pairwise disjoint rays. This ensures that infinite end-degree is realized by an actual infinite family of disjoint rays rather than merely by unbounded finite witnesses (Heuer, 2015). In the countable case, the grid theorem provides the canonical connected ray configuration for such an end; later work on Halin's end degree conjecture treats this as the $\aleph_0$-case of a more general program of “typical” ray configurations (Geschke et al., 2020).

2. Grid models and the choice of witness graph

Several grid-like graphs occur in the literature, and the precise target depends on the variant of the theorem. One common model is the hexagonal half-grid $H_{\mathrm{hex}}$, formalized as the graph on $\mathbb N \times \mathbb N$ with all vertical edges and horizontal edges $((n,m),(n+1,m))$ whenever $n+m$ is even. Its vertical rays are the columns $R_n=((n,0),(n,1),(n,2),\dots)$ (Kurkofka et al., 2021).

Another common witness is the hexagonal quarter grid $H^\infty$. Recent expositions note that $n$0 and the infinite square grid $n$1 each contain the other as a topological minor, so the particular choice of hexagonal versus square witness is not essential at the level of topological-minor existence (Chern, 12 Sep 2025). In coarse graph theory, however, the preferred object is often the square half-grid

$n$2

with edges between vertices at $n$3-distance $n$4. This $n$5 is the target of the coarse Halin theorem because its large-scale geometry is more directly suited to asymptotic-minor statements (Albrechtsen et al., 17 Jul 2025).

The distinction between half-grid and full grid is structurally significant. Halin's classical theorem detects “one-sided” thickness through a half-grid. By contrast, full-grid phenomena require stronger cyclic organization. Heuer's characterization of $n$6 grid minors shows that the full square grid is controlled not merely by infinitely many equivalent rays, but by bundles and consistent $n$7-bundles, which encode a two-sided ring structure around the rays (Heuer, 2015). Under quasi-transitivity, there is also a full hexagonal-grid strengthening: a locally finite quasi-transitive graph that is not quasi-isometric to a tree contains a subdivision of the full hexagonal lattice $n$8 (Georgakopoulos et al., 2024).

3. Strengthenings of the classical theorem

A major refinement of Halin's theorem gives control over which rays serve as the grid columns. The theorem of Kurkofka, Melcher, and Pitz states that for every infinite family $n$9 of pairwise disjoint equivalent rays in a graph $n \in \mathbb N$0, there exists a subdivision of the hexagonal half-grid such that every vertical ray of the subdivision is an element of $n \in \mathbb N$1 (Kurkofka et al., 2021). Classical Halin only guarantees some half-grid with columns in the end; the strengthening guarantees a half-grid using a prescribed infinite family of rays.

The proof strategy in that strengthening is based on a single ray $n \in \mathbb N$2 meeting every ray in $n \in \mathbb N$3 infinitely often and an auxiliary multigraph $n \in \mathbb N$4. The vertices of $n \in \mathbb N$5 are the rays in $n \in \mathbb N$6, and edges encode maximal $n \in \mathbb N$7-segments whose endpoints lie on two rays of $n \in \mathbb N$8. Passing to the spanning subgraph $n \in \mathbb N$9 obtained by deleting edges of finite multiplicity yields a combinatorial dichotomy: either $\aleph_0$0 has an infinite component, from which a half-grid subdivision is extracted directly, or one performs an inductive construction of columns and horizontal connections using vertices of infinite degree in suitable $\aleph_0$1 (Kurkofka et al., 2021).

A later short proof compresses the classical argument further. It proves a “star case” embedding lemma: if a graph contains a central ray $\aleph_0$2, disjoint side rays $\aleph_0$3, and infinitely many $\aleph_0$4–$\aleph_0$5 linkages with a finite-intersection condition, then the graph contains a topological $\aleph_0$6. To reduce a thick end to this configuration, the proof builds an auxiliary graph whose vertices are rays in the end and applies the star–comb lemma. The resulting star or comb of rays yields the grid embedding (Chern, 12 Sep 2025). This formulation makes explicit that the essential combinatorial content of Halin's theorem is the extraction of a suitable ray system with controlled linkage.

4. Coarse Halin theory and asymptotic minors

The classical theorem is not quasi-isometry invariant: a graph may contain a half-grid as a minor while collapsing its large-scale geometry. To address this, recent work replaces ordinary minors by $\aleph_0$7-fat, asymptotic, diverging, and ultra-fat minors. A model $\aleph_0$8 of a graph $\aleph_0$9 in $H_{\mathrm{hex}}$0 is $H_{\mathrm{hex}}$1-fat if distinct branch sets and branch paths are pairwise at distance at least $H_{\mathrm{hex}}$2, except for the obvious incidences between a branch set and a branch path corresponding to an incident edge. One writes $H_{\mathrm{hex}}$3 if $H_{\mathrm{hex}}$4 contains a $H_{\mathrm{hex}}$5-fat model of $H_{\mathrm{hex}}$6 for every $H_{\mathrm{hex}}$7 (Albrechtsen et al., 17 Jul 2025).

The main coarse Halin theorem states: if $H_{\mathrm{hex}}$8 is one-ended and locally finite, and if the disjoint union $H_{\mathrm{hex}}$9 of countably many rays is an asymptotic minor of $\mathbb N \times \mathbb N$0, then the square half-grid $\mathbb N \times \mathbb N$1 is also an asymptotic minor of $\mathbb N \times \mathbb N$2. More quantitatively, if $\mathbb N \times \mathbb N$3 for an end $\mathbb N \times \mathbb N$4, then $\mathbb N \times \mathbb N$5. Under stronger separation hypotheses on rays $\mathbb N \times \mathbb N$6, namely $\mathbb N \times \mathbb N$7, one obtains an ultra-fat half-grid minor $\mathbb N \times \mathbb N$8 (Albrechtsen et al., 17 Jul 2025).

The significance of this result is that asymptotic minors are quasi-isometry invariant, so the half-grid becomes a coarse prototype for thick-end behavior rather than merely a minor-theoretic one. The proof combines the prescribed-ray version of Halin's theorem with a geometric thickening-and-contraction procedure. One first uses a strengthening of Halin's theorem to obtain a subdivision of the hexagonal half-grid on selected rays, then passes through an auxiliary graph in which balls around the rays are contracted so that horizontal segments stay far from non-incident vertical rays, and finally contracts the resulting subdivision to a $\mathbb N \times \mathbb N$9-fat or ultra-fat model of $((n,m),(n+1,m))$0 (Albrechtsen et al., 17 Jul 2025).

The same paper extends the conclusion beyond the locally finite setting, but only with additional hypotheses. In particular, the relevant family of rays must consist of pairwise separated fat rays that cannot be separated from one another by removing finitely many bounded-radius balls. These assumptions are necessary: the paper gives explicit counterexamples showing that asymptotic copies of $((n,m),(n+1,m))$1 do not force even a $((n,m),(n+1,m))$2-fat half-grid minor in general non-locally-finite graphs, and that the constant $((n,m),(n+1,m))$3 in the quantitative theorem is sharp (Albrechtsen et al., 17 Jul 2025).

5. Quasi-transitivity, full grids, and quasi-isometric structure

Symmetry upgrades the classical half-grid conclusion. The full Halin grid theorem proves that every locally finite quasi-transitive graph that is not quasi-isometric to a tree contains a subdivision of the full hexagonal grid $((n,m),(n+1,m))$4. Equivalently, in this class, thick ends force the full hexagonal lattice rather than merely a half-plane of it. Two recorded corollaries are that every one-ended locally finite vertex-transitive graph contains a subdivision of $((n,m),(n+1,m))$5, and that in a locally finite quasi-transitive graph, the presence of a subdivision of the half-grid already implies the presence of a subdivision of the full grid (Georgakopoulos et al., 2024).

The proof is modular. In the planar one-ended case, bounded degree and bounded co-degree yield a diverging double ray that separates the plane, from which one recursively builds diverging layers linked by combs until the full hexagonal lattice emerges. For general quasi-transitive graphs, one passes to canonical tree-decompositions with finite adhesion and planar or bounded-treewidth torsos; a thick end must concentrate in a planar torso, where the planar argument applies (Georgakopoulos et al., 2024). This establishes a close connection between thick ends, quasi-transitivity, and non-tree-like large-scale geometry.

The coarse theory yields parallel consequences. Every one-ended, quasi-transitive, locally finite graph contains the square half-grid as an ultra-fat minor, hence as an asymptotic minor, and moreover admits a diverging minor-model of $((n,m),(n+1,m))$6. This applies in particular to all locally finite Cayley graphs of one-ended finitely generated groups, giving an affirmative answer to a problem of Georgakopoulos and Papasoglu (Albrechtsen et al., 17 Jul 2025). In accessible connected quasi-transitive locally finite graphs, the following are equivalent: having a thick end, containing $((n,m),(n+1,m))$7 as an ultra-fat minor, containing $((n,m),(n+1,m))$8 as an asymptotic minor, containing a diverging $((n,m),(n+1,m))$9-minor, and not being quasi-isometric to a tree (Albrechtsen et al., 17 Jul 2025).

These results place Halin-type grid phenomena squarely inside geometric group theory and coarse graph theory. The half-grid is no longer just a minor-theoretic obstruction but a quasi-isometry-invariant marker of non-tree-like behavior, while the full hexagonal grid becomes the rigid witness in sufficiently symmetric locally finite settings (Georgakopoulos et al., 2024).

6. Boundaries, higher-cardinal ends, and generalizations

The theorem has sharp limits. In the coarse setting, there exists a one-ended non-locally-finite graph $n+m$0 with $n+m$1 but $n+m$2, showing that local finiteness cannot simply be discarded. There is also, for every $n+m$3, a one-ended locally finite graph with $n+m$4 but $n+m$5, proving the sharpness of the quantitative threshold $n+m$6 (Albrechtsen et al., 17 Jul 2025).

For full square-grid minors, Heuer showed that the correct invariant is bundle structure rather than mere end-degree. A graph contains a $n+m$7 grid minor if and only if it has arbitrarily large bundles in one end; equivalently, if and only if it contains an $n+m$8-bundle, a consistent $n+m$9-bundle, or a subdivision of the Dartboard. Excluding a full grid minor is characterized by the existence of a bundle-narrow tree-decomposition into finite parts distinguishing all ends (Heuer, 2015). This clarifies the gap between Halin's half-grid theorem and two-sided full-grid structure.

The attempt to generalize Halin's theorem from countable end-degree to arbitrary cardinals leads to Halin's end degree conjecture. For regular uncountable $R_n=((n,0),(n,1),(n,2),\dots)$0, the conjectural typical configuration is a $R_n=((n,0),(n,1),(n,2),\dots)$1-star of rays. The current picture is mixed: the conjecture fails at $R_n=((n,0),(n,1),(n,2),\dots)$2, holds for $R_n=((n,0),(n,1),(n,2),\dots)$3, fails at $R_n=((n,0),(n,1),(n,2),\dots)$4, and is independent of ZFC at $R_n=((n,0),(n,1),(n,2),\dots)$5 for $R_n=((n,0),(n,1),(n,2),\dots)$6 (Geschke et al., 2020). At degree $R_n=((n,0),(n,1),(n,2),\dots)$7, a later paper gives a consistent example of an end of degree $R_n=((n,0),(n,1),(n,2),\dots)$8 with no ray graph, while proving in ZFC that every end of degree $R_n=((n,0),(n,1),(n,2),\dots)$9 still satisfies the weaker property $H^\infty$0: every countable subfamily of disjoint rays extends to a countable ray-graph witness (Aurichi et al., 2024). This suggests that the $H^\infty$1-grid phenomenon survives at $H^\infty$2 only as a countable-window principle, not as a global grid or star.

There is also a directed analogue. For an infinite family of pairwise disjoint equivalent out-rays or in-rays in a digraph, one obtains a subdivision of either a bidirected quarter-grid or a dominated directed quarter-grid whose vertical rays come from the given family. An analogous theorem holds for necklaces, the strongly connected directed counterparts of rays used in directed-end theory (Reich, 2024). In the directed setting, unlike the undirected one, two distinct quarter-grid types are genuinely necessary: neither subsumes the other while preserving the prescribed vertical family (Reich, 2024).

Taken together, these developments show that Halin's Grid Theorem is both rigid and extensible. It is rigid in the countable undirected case, where thick ends force a canonical half-grid structure; it is extensible in that the same idea admits prescribed-ray, coarse, symmetric, full-grid, higher-degree, and directed formulations. The central invariant throughout is the organization of infinitely many disjoint rays inside an end, and the various modern theorems can be viewed as progressively finer descriptions of how that organization manifests itself in the ambient graph.