Groups with a finite Busemann boundary are virtually cyclic

Ariel Yadin; Corentin Bodart; Liran Ron-George

REVIEW 1 major objections 3 minor 16 references

Groups with a finite Busemann boundary are virtually cyclic

T0 review · 1 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A finitely generated group with a Cayley graph that has only finitely many Busemann points must be virtually cyclic.

desk verdict The paper proves the missing implication that a single Cayley graph with finitely many Busemann points forces virtual cyclicity, with a neat new annihilator tool—but one lemma leans on an unpublished corollary in the authors' own preprint, so the proof isn't fully self-contained. read the letter →

arxiv 2511.20495 v2 pith:5IAR6UZ2 submitted 2025-11-25 math.GR math.MG

Corentin Bodart , Liran Ron-George , Ariel Yadin This is my paper · ORCID

classification math.GRmath.MG MSC 20F65

keywords metric-functionalboundaryBusemannpointsvirtuallycyclicgroupsCayleygraphsannihilatorsubgrouphorofunctiongeometricgrouptheory

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proves that a finitely generated infinite group is virtually cyclic if and only if some Cayley graph of the group has only finitely many Busemann points—the boundary points that arise as limits of Busemann functions along geodesics. This completes a chain of equivalences: having a finite metric-functional boundary for every Cayley metric, having a Cayley graph with finitely many Busemann points, and being virtually cyclic are all the same property. The new, difficult direction shows that even a single Cayley graph with a finite Busemann set forces the group to contain a finite-index copy of the integers. The proof introduces the annihilator of a boundary set, the subgroup of elements that evaluate to zero on all its functions, and shows that finiteness of the Busemann set makes the annihilator finite. From there the group's kernel on the finite boundary is shown to be virtually abelian, and a previously settled virtually abelian case completes the argument.

What carries the argument

The annihilator of a subset L of the metric-functional boundary, N_L = {x∈G : h(x)=0 for all h∈L}, which is a subgroup when L is invariant. For Cayley metrics the annihilator of the Busemann set equals the annihilator of the whole metric-functional boundary (Theorem 1.4). The proof's geometric engine is Lemma 5.1, which constructs long finite geodesics whose midpoint values stay uniformly bounded on all Busemann functionals; iterating this over the annihilator yields the finiteness of N and hence the virtually abelian reduction.

What would settle it

Find a finitely generated, infinite, non-virtually-cyclic group G and a Cayley metric d_S whose Busemann set is finite; that would refute Theorem 1.1. Alternatively, exhibit a group satisfying the hypotheses of Lemma 4.3—finite Busemann set and an action kernel with a Z^2 quotient—to refute the lifting step on which the proof depends.

Watch

Extended reading notes

Core claim

Theorem 1.1 asserts that for a finitely generated infinite group G the following are equivalent: (1) every Cayley metric has finite metric-functional boundary; (2) some Cayley metric has finitely many Busemann points; (3) G is virtually Z. The new implication is (2)⇒(3). The proof shows the annihilator N of the Busemann set is a finite subgroup (Lemma 1.5), and the kernel K of the action on the finite Busemann set then has K/(K∩N) embedded in Z^n, so K is virtually abelian; applying Theorem 1.2 to this virtually abelian group yields virtual cyclicity of G.

Load-bearing premise

The proof of Lemma 4.3 assumes a companion result, not proved here, that a finitely generated group with a quotient that is virtually Z^d for d≥2 has infinitely many Busemann points in every Cayley metric; if that lifting statement fails, the contradiction that yields virtual cyclicity collapses.

Editorial extensions

If this is right

The property 'has a Cayley graph with finitely many Busemann points' is a group invariant: if it holds for one generating set, the group is virtually cyclic and therefore it holds for every generating set.
Every finitely generated group that is not virtually cyclic has infinitely many Busemann points in any Cayley graph.
The metric-functional boundary is finite for all Cayley metrics exactly when the group is virtually cyclic; the earlier partial results and this paper now form one characterization.
The annihilator of the Busemann set coincides with the annihilator of the full metric-functional boundary, and annihilators of invariant subsets are locally finite for proper integer-valued left-invariant metrics, giving a new algebraic handle on boundaries.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The proof's reliance on an external quotient-lifting result suggests a natural next step: a self-contained proof of that step would make Theorem 1.1 independent of companion preprints, and a counterexample would localize exactly where the argument fails.
Question 1.9—can a Cayley metric have an infinite annihilator?—is concrete and testable: adapting the non-Cayley construction of Example 3.3 to a genuine Cayley metric would answer it, and a negative answer would strengthen the link between finite Busemann sets and finite annihilators.
The annihilator construction transfers naturally to other boundaries (e.g., horofunction boundaries of more general metric spaces): one could test whether finiteness of the boundary is always witnessed by a finite annihilator subgroup.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The new theorem is real: if one Cayley graph on a finitely generated group has finitely many Busemann points, the group is virtually cyclic, completing the equivalence. The annihilator notion is the right tool for this, and the proof that the annihilator of the full metric-functional boundary equals that of the Busemann points is clean. The local-finiteness result (Prop 1.6) and the examples are also nice side contributions. The proof of Theorem 4.1 for virtually abelian groups is worked out in detail and looks correct to me; the discussion around Lemma 4.2 is careful.

The main soft spot is in Lemma 4.3. To rule out dim V ≥ 2, the proof uses that G/[K,K] is virtually Z^d for d≥2 and then cites Corollary 1.11 from the authors' own arXiv preprint [2] to conclude the Busemann boundary is infinite. That corollary is not proved here nor stated as an explicit assumption. Theorem 4.1 only covers groups containing a finite-index Z^d subgroup, not groups with merely a virtually-Z^d quotient. So this is a genuine gap in the manuscript as written: the contradiction in Lemma 4.3 depends entirely on that unpublished lifting result. If the corollary is true, the proof goes through; if not, the main theorem collapses. The authors should either prove it, supply a published reference, or mark it as a hypothesis.

Everything after Lemma 4.3—Lemma 5.1, Lemma 1.5, and the final argument for Theorem 1.1—seems sound once that assumption is granted. The inequality in Lemma 5.1 is a bit awkwardly stated but reconstructs as the reader notes; not a real problem.

The paper is aimed at people working on horofunction boundaries and geometric group theory. The result is worth knowing and the annihilator construction may be reused. I'd recommend sending it to peer review, but with a clear request to resolve the dependency on [2] before acceptance. It deserves a serious referee, but it shouldn't be published in its current form.

Referee Report

1 major / 3 minor

Summary. The paper proves Theorem 1.1: a finitely generated, infinite group G has a Cayley metric with finitely many Busemann points if and only if G is virtually Z. The new implication is (2)⇒(3): the existence of a single Cayley graph with finite Busemann boundary forces virtual cyclicity. The proof introduces 'annihilators' N_L for invariant subsets L of the metric-functional boundary, shows that for Cayley metrics the annihilator of the Busemann boundary equals the annihilator of the full boundary (Theorem 1.4), proves that the annihilator is locally finite (Proposition 1.6), and then uses a convex-hull construction to prove Theorem 4.1: any group with a finite-index Z^d (d≥2) subgroup has infinitely many Busemann points for every Cayley metric. Lemma 4.3 then shows that, under finiteness of the Busemann boundary, two Busemann points are proportional on the kernel K of the action, with proportionality constant ±1. Combining these, Lemma 1.5 proves the annihilator N is finite, and Theorem 1.1 follows by reducing to the virtually abelian case and applying Theorem 1.2.

Significance. If the main theorem is correct, it completes the characterization of finitely generated groups with finite metric-functional boundary, a question motivated by Karlsson's metric-functional program. The annihilator framework is a genuinely new and potentially useful tool for studying Busemann boundaries. The proof is largely self-contained and contains an elegant convex-hull argument (Theorem 4.1) that is a nice contribution in its own right. However, a load-bearing step in Lemma 4.3 depends on an unproved, unpublished corollary from a co-authored preprint. This is currently the only serious obstacle to immediate acceptance.

major comments (1)

[§4, Lemma 4.3 (the sentence 'Combining Theorem 4.1 and Corollary 1.11 in [2]')] The proof that dim V ≥ 2 leads to a contradiction uses Corollary 1.11 of [2], which is an unpublished arXiv preprint (Bodart–Tashiro). The corollary is not stated in this paper, nor is its proof provided, and it is not a standard result in the literature. The step is genuinely load-bearing: it is the only place where the case dim V ≥ 2 is ruled out, and it is needed for Lemma 5.1 and Lemma 1.5, and ultimately for the new direction (2)⇒(3) of Theorem 1.1. The authors should either (a) include a self-contained proof of the needed lifting statement (e.g., if G/[K,K] is virtually Z^d with d≥2 then |∂_b(G,d_S)|=∞ for every Cayley metric), or (b) if the preprint [2] is accepted, give a precise citation to the published version. Without this, the proof of the main theorem is incomplete.

minor comments (3)

[§4, Step I of Theorem 4.1] In the induction step, when z = 1, the expression ξ(z)/|z|_S is undefined; the convex-combination argument implicitly uses 0 ∈ P. This should be handled explicitly (e.g., say that if z=1 the term is omitted).
[§5, proof of Theorem 1.1] The phrase 'K/(K∩N) is an infinite subgroup of Z^n' is imprecise: K/(K∩N) is isomorphic to a subgroup of Z^n, not literally a subgroup. The intended meaning is clear.
[§1.3, Example 1.11] The notation 'S = F S_1 F' could be clarified: it should be stated that F is finite and the product is understood as the set {f s f' : f,f'∈F, s∈S_1}, which is finite and symmetric. The argument is correct.

Circularity Check

1 steps flagged · score 4.0 of 10

Lemma 4.3's key contradiction is imported from co-authored preprint [2] via Corollary 1.11, leaving the central new direction dependent on an unproved self-citation.

self citation load bearing [Section 4, Lemma 4.3 (proof, paragraph after defining V=span φ(K))]
"It follows that G/[K,K]≥f.i. K/[K,K]≃Z^d×T with d≥2 and T finite, that is, G has a quotient which is virtually-Z^d for d≥2. Combining Theorem 4.1 and Corollary 1.11 in [2], we get that |∂_b(G,d_S)|=∞, a contradiction!"

The contradiction eliminating the dim V≥2 case rests entirely on Corollary 1.11 of the Bodart–Tashiro preprint [2], which is not proved or even stated in this paper. Since [2] shares an author and is used to pass from a virtually-Z^d quotient to an infinite Busemann boundary, the key step is imported by self-citation rather than derived. If that corollary fails, Lemma 4.3's proof has no alternative argument, and the downstream Lemmas 5.1 and 1.5 — and hence the new direction (2)⇒(3) — collapse. This is not a definitional identity, but it is a load-bearing dependency on an unverified co-authored result.

full rationale

The proof of the new implication (2)⇒(3) is largely original: the annihilator machinery, Lemma 1.5, Lemma 5.1, and the final reduction to the virtually abelian case are derived internally. However, the proof of Lemma 4.3 has a load-bearing citation. After constructing a quotient of G that is virtually Z^d with d≥2, the paper does not prove that such a quotient forces infinitely many Busemann points; it invokes Corollary 1.11 of the co-authored preprint [2]. This is the only step ruling out dim V≥2 in Lemma 4.3, and Lemma 4.3 is needed for Lemma 5.1 and Lemma 1.5, which in turn are used to prove Theorem 1.1. Thus a central node of the derivation is supported by a self-citation to an unpublished, non-machine-checked result. This is not a definitional equivalence or a fitted-parameter prediction, and the surrounding argument has independent mathematical content; if [2] is treated as a valid external theorem, the paper is essentially non-circular. But on the evidence of the present text, the dependency is real and load-bearing, so the score is moderate rather than 0.

Assumptions & free parameters 0 free parameters · 8 assumptions · 0 invented entities

Free parameters: none — this is a proof-based paper with no data fitting. Axioms: the paper relies on standard mathematics (pointwise compactness, convex separation) and on several domain-specific results from the prior literature, including two co-authored works: the published [12] (Ron-George–Yadin) and the unpublished preprint [2] (Bodart–Tashiro). The annihilator is a definition, not a postulated entity, so invented_entities is empty.

assumptions (8)

domain assumption Countable, finitely generated groups; proper left-invariant metrics; Cayley graph word metrics are proper and geodesic.
Stated in Section 1.1 and used throughout.
domain assumption [12, Proposition 2.2] — an infinite sequence y_n with |y_n|→∞ yields a boundary point as limit of b_{y_n}; Busemann points are unbounded.
Used in Lemma 3.1 and Lemma 4.3 to characterize membership in N and to show φ is nontrivial.
domain assumption [12, Lemma 3.4] — a 1-Lipschitz homomorphism f on H with f(w)=|w| gives a geodesic γ with γ_{n|w|}=w^n.
Used in Lemma 4.2 to construct geodesics along w^n.
domain assumption [15, Proposition 2.1] — two geodesics converging to the same Busemann point have a third geodesic intersecting both infinitely often.
Used in Lemma 4.2 to force y^{-1}z ∈ ⟨x⟩.
domain assumption [2, Corollary 1.11] — a group with a quotient that is virtually Z^d (d≥2) has infinite Busemann boundary for every Cayley metric.
Used in Lemma 4.3 to get the contradiction dim V = 1; load-bearing and from an unpublished co-authored preprint.
domain assumption [12] and [14] — virtually-Z groups have finite metric-functional boundary for all Cayley metrics.
Used for the converse direction in Theorem 1.1.
standard math Convex hull separation: for a finite F⊂R^d and extreme point e∈conv(F), there exists a linear functional with φ(p)≤1 for all p∈P and φ(p)=1 iff p=e.
Used in Theorem 4.1, Step II, to construct the 1-Lipschitz homomorphism f.
standard math Compactness of {0,1}^G and pointwise convergence give limits of geodesic sequences (Lemma 2.2).
Used to extract infinite geodesics from growing finite geodesics.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Groups with a finite Busemann boundary are virtually cyclic." pith.science (2026). https://pith.science/paper/5IAR6UZ2

@misc{pith2026251120495,
  author       = {Pith},
  title        = {Pith review of: Groups with a finite Busemann boundary are virtually cyclic},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5IAR6UZ2}},
  note         = {Machine review of arXiv:2511.20495}
}

read the original abstract

This note is a continuation of the study of the relationship between the geometry of Cayley graphs and the size of its metric-functional boundary. We show that, if there exists a Cayley graph with finitely many Busemann points, then the underlying group is virtually cyclic. Together with previous works, this completes the full characterization of groups with finite metric-functional boundaries. The main new notion introduced is that of annihilators.

Figures

Figures reproduced from arXiv: 2511.20495 by the authors.

**Figure 1.** The Cayley graph of Z × Z/nZ, with N in red. Example 1.11 Consider a group G, a finite symmetric generating set S1 and a finite subgroup F. We take S = F S1F and prove that N = N∂(G,dS) ≥ F. For all f ∈ F and y ∈ G − F, we have dS(f, y) = dS(1, y). Indeed, if y = s1s2 . . . sℓ with si ∈ S, then f −1y = (f −1 s1)s2 . . . sℓ and f −1 s1 ∈ S, proving that dS [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗

**Figure 2.** Finite geodesics α (n) with endpoints in N limiting to an infinite geodesic β satisfying Equation (1). The next idea is to start from the first “bend” in each geodesics α (n) where Equation (1) is not locally satisfied, translate the remaining part back to 1 to get long finite geodesics that “start slowly with respect to γ∞” (see [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗

**Figure 3.** The geodesics α (n) , with their “bend” translated back to 1 so that Equation (1) fails from the start. Lemma 5.1 Let dS be a Cayley metric on a group G such that the set of Busemann points ∂b(G, dS) is finite. Recall that K = {x ∈ G : ∀ h ∈ ∂b(G, dS) , x.h = h} is the kernel of the action of G on ∂b(G, dS), and N = N∂b(G,dS) is the annihilator. Then, for any m > 0 and any ℓ satisfying m ≤ ℓ ≤ 2 m + 2 sup |x|S : x… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

16 extracted references · 3 linked inside Pith

[2]

Bodart and K

C. Bodart and K. Tashiro. Horofunctions on the Heisenberg and Cartan groups.arXiv preprint arXiv:2407.11943, 2024

arXiv 2024
[1]

Bader and V

U. Bader and V. Finkelshtein. On the horofunction boundary of discrete Heisenberg group.Geome- triae Dedicata, 208(1):113–127, 2020

2020
[3]

M. Develin. Cayley compactifications of abelian groups.Annals of Combinatorics, 6(3):295–312, 2002

2002
[4]

Fisher and S

N. Fisher and S. Nicolussi Golo. Sub-finsler horofunction boundaries of the heisenberg group.Anal- ysis and Geometry in Metric Spaces, 9(1):19–52, 2021

2021
[5]

M. Gromov. Hyperbolic manifolds, groups and actions. InRiemann surfaces and related topics: Proceedings of the 1978 Stony Brook Conference (State Univ. New York, Stony Brook, NY, 1978), volume 97, pages 183–213, 1981

1978
[6]

Kapovich and N

I. Kapovich and N. Benakli. Boundaries of hyperbolic groups.arXiv preprint math/0202286, 2002

arXiv 2002
[7]

Karlsson

A. Karlsson. Non-expanding maps and busemann functions.Ergodic Theory and Dynamical Sys- tems, 21(5):1447–1457, 2001

2001
[8]

Karlsson

A. Karlsson. From linear to metric functional analysis.Proceedings of the National Academy of Sciences, 118(28):e2107069118, 2021

2021

Show all 16 references

[9]

Karlsson

A. Karlsson. Hahn-Banach for metric functionals and horofunctions.Journal of Functional Analysis, 281(2):109030, 2021

2021
[10]

Karlsson

A. Karlsson. A metric fixed point theorem and some of its applications.Geometric and Functional Analysis, pages 1–26, 2024

2024
[11]

M. A. Rieffel. GroupC ∗-algebras as compact quantum metric spaces.Documenta Mathematica, 7:605–651, 2002

2002
[12]

Ron-George and A

L. Ron-George and A. Yadin. Groups with finitely many Busemann points.Groups, Geometry and Dynamics, 2023. arXiv:2305.02303

2023 arXiv
[13]

Ron-George and A

L. Ron-George and A. Yadin. Detecting virtual homomorphisms via Banach metrics.arXiv preprint arXiv:2408.11543, 2024

2024
[14]

Tointon and A

M. Tointon and A. Yadin. Horofunctions on graphs of linear growth.Comptes Rendus Mathematique, 354(12):1151–1154, 2016

2016
[15]

C. Walsh. The action of a nilpotent group on its horofunction boundary has finite orbits.Groups, Geometry, and Dynamics, 5(1):189–206, 2011

2011
[16]

Webster and A

C. Webster and A. Winchester. Busemann points of infinite graphs.Transactions of the American Mathematical Society, 358(9):4209–4224, 2006

2006

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [2]

Bodart and K

C. Bodart and K. Tashiro. Horofunctions on the Heisenberg and Cartan groups.arXiv preprint arXiv:2407.11943, 2024

arXiv 2024

[2] [1]

Bader and V

U. Bader and V. Finkelshtein. On the horofunction boundary of discrete Heisenberg group.Geome- triae Dedicata, 208(1):113–127, 2020

2020

[3] [3]

M. Develin. Cayley compactifications of abelian groups.Annals of Combinatorics, 6(3):295–312, 2002

2002

[4] [4]

Fisher and S

N. Fisher and S. Nicolussi Golo. Sub-finsler horofunction boundaries of the heisenberg group.Anal- ysis and Geometry in Metric Spaces, 9(1):19–52, 2021

2021

[5] [5]

M. Gromov. Hyperbolic manifolds, groups and actions. InRiemann surfaces and related topics: Proceedings of the 1978 Stony Brook Conference (State Univ. New York, Stony Brook, NY, 1978), volume 97, pages 183–213, 1981

1978

[6] [6]

Kapovich and N

I. Kapovich and N. Benakli. Boundaries of hyperbolic groups.arXiv preprint math/0202286, 2002

arXiv 2002

[7] [7]

Karlsson

A. Karlsson. Non-expanding maps and busemann functions.Ergodic Theory and Dynamical Sys- tems, 21(5):1447–1457, 2001

2001

[8] [8]

Karlsson

A. Karlsson. From linear to metric functional analysis.Proceedings of the National Academy of Sciences, 118(28):e2107069118, 2021

2021