Homology in Combinatorial Refraction Billiards

Colin Defant; Derek Liu

REVIEW 1 major objections 5 minor 17 references

Homology in Combinatorial Refraction Billiards

T0 review · 1 major / 5 minor · reviewed 2026-08-08 · deepseek-v4-flash

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

Pith's one-line read The paper proves a torus billiard system built from a graph expels every trajectory exactly when the graph is bipartite, and gives a partial classification of the opposite, ensnaring case.

desk verdict Clean new classification — expelling iff bipartite — with a correct proof; later sections are sketchier but the core is solid. read the letter →

arxiv 2502.06013 v2 pith:HRVNPB2T submitted 2025-02-09 math.CO

Colin Defant , Derek Liu This is my paper · ORCID

classification math.CO MSC 05C7520F5552C35

keywords combinatorialbilliardstorichyperplanearrangementsaffinesymmetricgroupensnaringgraphsexpellingbipartitewindingvectorsstonediagrams

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

This paper studies a billiard system on a torus encoded by a graph $G$: beams of light travel through an $(n-1)$-dimensional torus and bend whenever they cross hyperplanes whose adjacency pattern is $G$. Every trajectory is periodic, so it can be viewed as a closed loop, and the paper asks whether such loops are contractible. A graph is called ensnaring if all trajectories are contractible and expelling if none are; the first main result is that $G$ is expelling if and only if $G$ is bipartite. Ensnaring graphs turn out to be harder to classify, and the paper supplies a collection of necessary and sufficient conditions along with operations that build new ensnaring graphs from old ones.

What carries the argument

The load-bearing object is the stone diagram, a finite combinatorial picture of an orbit: each vertex of $G$ has a labeled replica on a cycle of length $n$, and a stone with a clockwise or counterclockwise orientation marks the active position. The map $\Theta_G$ swaps the two replicas next to the stone and either moves the stone (if the corresponding vertices are non-adjacent) or reverses its direction (if they are adjacent). For an orbit $\mathcal{O}$, the winding vector $\vec w_{\mathcal O}\in\mathbb{Z}^n$ records the net clockwise displacement of each replica after one full period, and Proposition 2.4 reduces the topological question to algebra: a trajectory is contractible exactly when this vector is the zero vector, because zero winding is equivalent to boundedness of every lift to the affine symmetric group. Ensnaring and expelling thus become properties of the winding vectors of a finite dynamical system.

What would settle it

Run the orbit of $\Theta_G$ for the triangle graph $K_3$ starting from the identity stone diagram $(\mathrm{id},1,1)$; the paper predicts the orbit has zero winding vector. Alternatively, run any orbit for an even cycle such as $C_4$, which the paper predicts always has nonzero winding vector; a single counterexample to either prediction would disprove the expelling-bipartite dichotomy.

Watch

Extended reading notes

Core claim

The central claim is a dichotomy: a graph $G$ is expelling, meaning no billiard trajectory in the toric system is contractible, if and only if $G$ is bipartite (Theorem 3.1). In the bipartite direction, the stone-diagram model shows that two replicas belonging to opposite parts of the bipartition always wind in opposite directions, so the winding vector cannot be zero. In the converse direction, an induced odd cycle is used to build a periodic orbit whose replicas never cross a particular edge of the surrounding cycle, forcing the winding vector to vanish and the trajectory to contract. The paper then studies ensnaring graphs, where all trajectories are contractible: complete graphs $K_n$ for $n\ge 3$ are ensnaring, a cycle $C_n$ is ensnaring exactly for odd $n$, wedging two ensnaring graphs at a vertex preserves ensnaring, and a disjoint union is ensnaring exactly when both factors are ensnaring and neither is revolutionary, meaning no orbit's stone has nonzero winding. It also proves a complement reduction: a graph with $n$ vertices is ensnaring exactly when, for every connected component $C$ of its complement, the $n$-vertex complement of $C$ is ensnaring.

Load-bearing premise

The classification rests on identifying contractibility of a continuous toric trajectory with vanishing of the discrete winding vector; if a continuous loop could fail to contract while its discretization had zero winding, or vice versa, the theorems would no longer follow.

Editorial extensions

If this is right

Every bipartite graph is expelling, and every expelling graph is bipartite, so the topological behavior of all trajectories in the expelling case is decided by a single classical graph invariant.
Complete graphs with at least three vertices are ensnaring: the stone never leaves its initial position, so no replica crosses the cycle edge opposite the stone, forcing zero winding.
A cycle graph $C_n$ is ensnaring exactly when $n$ is odd; even cycles are expelling by bipartiteness, while odd cycles force all replicas to share the same winding number, which must be zero.
Wedging two ensnaring graphs at a vertex preserves ensnaring, whereas a disjoint union is ensnaring exactly when both factors are ensnaring and neither is revolutionary.
The complement reduction says ensnaring can be checked componentwise on the complement, and it implies, for instance, that a graph whose complement has a clique of size at least two as a connected component is not ensnaring.

Reading between the lines

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

The bipartite dichotomy suggests a broader principle: for the mirror-and-refraction generalization sketched in Section 9, expelling may correspond to contracting all reflection edges and then asking whether the quotient is bipartite; testing this on materialized cycles would be a direct next step.
The local obstruction in Theorem 7.1 indicates that non-ensnaring can be caused by small induced configurations, so it is reasonable to ask whether ensnaring admits a forbidden-induced-configuration characterization rather than only a global one.
The parity condition in the classification of complements of complete bipartite graphs points toward a general rule in which the ensnaring status of an $n$-vertex complement depends only on the parity of $n$, as Conjecture 8.3 proposes.
The winding-vector criterion is effectively a discrete homology invariant for periodic trajectories, so the paper's result offers a purely finite way to certify contractibility or non-contractibility in a continuous torus billiard system.

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 headline is that this paper gives a genuinely new classification: a graph is expelling if and only if it is bipartite. The ensnaring/expelling dichotomy is a real addition to the combinatorial billiards literature, and the stone-diagram model reduces a topological question about loop contractibility to a purely combinatorial winding vector condition. That reduction (Proposition 2.4) is clearly argued. The complement-component theorem and the wedge and disjoint-union constructions give the ensnaring family enough structure to be worth studying.

I checked the stress-test note about Proposition 3.2. It doesn't hold up. The alleged counterexample lists a four-state orbit in reverse order. In the forward dynamics, both edge swaps between the X-vertex and the Y-vertex have the X-vertex moving clockwise, exactly as the proof claims. More generally, the proof's monotonicity claim is correct: when the coin is on an X-vertex, the stone points clockwise; when on a Y-vertex, it points counterclockwise; and in either case the X-replica moves clockwise during a swap with a Y-replica, while the Y-replica moves counterclockwise. So the proof mechanism is valid. The stress-test's own admission that the winding vector is nonzero shows it isn't a counterexample to the theorem.

The real soft spots are later. Theorem 5.5, Proposition 6.1, Theorem 6.3, and Theorem 8.6 all lean on phrases like "straightforward," "similar to the proof of Lemma 5.1," or "the same argument." I don't see an actual error, but these are places where a referee should demand details. The paper is honest about this: Section 9.2 labels its generalizations as sketches and conjectures, and the open problems are clearly separated from proven results.

The citation pattern is fine. The paper depends on [1] for the billiard model and one tree lemma, but that's the natural predecessor, not a hidden assumption. Self-citation here is appropriate since the prior paper actually does the heavy lifting for the model.

Who is this for? People working in dynamical algebraic combinatorics and anyone who likes clean graph invariants controlling a dynamical system. The main theorem is memorable and the proof of the main theorem is self-contained enough. I would send it to a serious referee. If I were the referee, my main request would be to expand the compressed proofs in Sections 5, 6, and 8.

Referee Report

1 major / 5 minor

Summary. The paper introduces a topological perspective on the toric combinatorial refraction billiards of Adams, Defant, and Striker. For a graph G, trajectories are periodic loops in an (n−1)-torus, and the paper defines G to be ensnaring if all such loops are contractible and expelling if none is contractible. The central result is Theorem 3.1, which characterizes expelling graphs as exactly the bipartite graphs. The paper then studies ensnaring graphs: complete graphs and odd cycles are ensnaring, wedges of ensnaring graphs are ensnaring, disjoint unions are governed by a new notion of revolutionary graphs, certain wedges of complete graphs with trees are ensnaring, and several necessary and sufficient conditions are given in terms of complements. The final section collects conjectures and a generalization to graphs with both reflection and refraction edges.

Significance. The bipartite characterization in Theorem 3.1 is a crisp and attractive result: it reduces a topological property of all toric refraction billiard trajectories of a graph to a single standard graph invariant. Proposition 2.4, which reduces contractibility to vanishing of the winding vector in the stone-diagram model, is a useful bridge between the continuous and combinatorial settings. The paper also provides explicit orbit-level constructions for many ensnaring and non-ensnaring examples, and the later complement theorems give a substantial family of dense ensnaring graphs. The proofs are largely self-contained, and the main theorem is supported by a clear geometric mechanism. The stress-test concern about Proposition 3.2 does not land: the proposed counterexample is obtained by traversing the orbit in the reverse direction and therefore does not contradict the forward-time monotonicity claim used in the proof.

major comments (1)

[§3, Proposition 3.2] The proof's final implication, 'This immediately implies that x and y have different winding numbers,' is terse. It is mathematically correct, but it relies on the standard fact that the difference of the winding numbers of two replicas equals the net number of times one passes the other; I recommend stating this explicitly. I also checked the alleged counterexample in the stress test. For the graph with edge {1,2}, bipartition X={1}, Y={2,3}, and the starting state (0,1,2) with the stone pointing clockwise and coexisting with replica 1, the forward orbit under Θ_G is (0,1,2) → (1,0,2) → (1,2,0) → (2,1,0) → (0,1,2), not the reversed sequence (0,1,2) → (2,1,0) → (1,2,0) → (1,0,2) → (0,1,2). In the forward orbit, both swaps of replicas 1 and 2 move replica 1 clockwise and replica 2 counterclockwise. Thus the proposed counterexample does not invalidate the proof.

minor comments (5)

[§3, Proposition 3.2] The invariant that the stone points clockwise when the coin is on a vertex in X and counterclockwise when the coin is on a vertex in Y would be clearer if the authors noted that non-adjacent swaps preserve the coin vertex and the stone orientation, while adjacent swaps change both; this is the whole content of the claim 'This implies...'.
[§5, Theorem 5.5] There is a typo in the proof: 'we may asssume n1 > 1' should read 'we may assume n1 > 1'.
[§8, Theorem 8.6] The case analysis in the proof of Theorem 8.6 is quite dense, especially the m-odd case; a table or figure tracking the bridged-edge orientations and the positions of a1 and a2 would improve readability.
[§2] The footnote attached to 'metalenses' is numbered with a stray '1' in the displayed text; the formatting should be cleaned up.
[§6, Theorem 6.3] The proof of Theorem 6.3 is compressed, particularly the sentence 'similar to the proof of Proposition 6.1, we know D′ can be obtained from taking D, swapping va and vc, and then taking a rotation'; adding a short justification or an illustrative figure would help the reader verify this key symmetry step.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the expelling/bipartite theorem is derived from the billiard model and winding-vector criterion rather than assumed.

full rationale

The paper's central claim, Theorem 3.1 (expelling if and only if bipartite), is not obtained by assuming its conclusion. Section 2 derives Proposition 2.4, which identifies contractibility of a toric billiard trajectory with vanishing winding vector, by lifting the closed trajectory to the Euclidean alcove path and showing that the displacement after one period is exactly the negative winding vector (Bu_lP = Bu_l - w_O). This is a substantive bridge from the continuous/topological setup to the combinatorial stone-diagram model, not a definitional equivalence imposed in the statement of the theorem. Proposition 3.2 then proves that bipartite graphs are expelling by analyzing stone orientation and replica motion, while Proposition 3.3 proves the converse by constructing a pleasant stone diagram on an odd cycle and arguing that its winding vector vanishes. Both arguments operate on the combinatorial model established independently in Section 2 rather than invoking Theorem 3.1. The paper does rely on the authors' prior article [1] for the original refraction-billiard framework, the definition of Theta_G, the stone-diagram convention, and a tree lemma used in Section 6; these are background tools or auxiliary results, and they are parameter-free lemmas whose assumptions do not include the present target theorem. No fitted parameter is later renamed as a prediction, and no uniqueness theorem from the authors' own work is used to forbid alternatives. The classification theorem therefore has independent mathematical content, and the derivation does not reduce to its inputs by construction.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

This is a pure mathematics paper with no data fitting and no new physical entities. The central proofs rely on standard Coxeter theory, the prior billiard discretization from [1], and one cited lemma about trees. The new combinatorial notions such as stone diagrams, winding vectors, and revolutionary graphs are internal definitions, not fitted parameters or independent physical postulates.

assumptions (3)

standard math H_n is the Coxeter arrangement of the affine symmetric group eS_n, with the stated faithful action on V.
Section 2 uses standard Coxeter theory to identify alcoves with group elements and to set up the torus quotient.
domain assumption The map Theta_G and the stone and coin diagrams from [1] faithfully discretize the continuous toric refraction billiards, and contractibility of a loop is detected by the winding vector.
This is the bridge between geometry and combinatorics used throughout the paper. Proposition 2.4 is argued in the text, but the underlying billiard discretization is inherited from prior work.
domain assumption In any orbit of a tree, every replica swaps past every other replica (Lemma 5.1 of [1]).
This cited lemma is used without proof in Corollary 6.2 and Theorem 6.3 to control orbits that enter trees wedged to a complete graph.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Homology in Combinatorial Refraction Billiards." pith.science (2026). https://pith.science/paper/HRVNPB2T

@misc{pith2026250206013,
  author       = {Pith},
  title        = {Pith review of: Homology in Combinatorial Refraction Billiards},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HRVNPB2T}},
  note         = {Machine review of arXiv:2502.06013}
}

abstract

Given a graph $G$ with vertex set $\{1,\ldots,n\}$, we can project the graphical arrangement of $G$ to an $(n-1)$-dimensional torus to obtain a toric hyperplane arrangement. Adams, Defant, and Striker constructed a toric combinatorial refraction billiard system in which beams of light travel in the torus, refracting (with refraction coefficient $-1$) whenever they hit one of the toric hyperplanes in this toric arrangement. Each billiard trajectory in this system is periodic. We adopt a topological perspective and view the billiard trajectories as closed loops in the torus. We say $G$ is ensnaring if all of the billiard trajectories are contractible, and we say $G$ is expelling if none of the billiard trajectories is contractible. Our first main result states that a graph is expelling if and only if it is bipartite. We then provide several necessary conditions and several sufficient conditions for a graph to be ensnaring. For example, we show that the complement of an ensnaring graph cannot have a clique as a connected component. We also discuss ways to construct ensnaring graphs from other ensnaring graphs. For example, gluing two ensnaring graphs at a single vertex always yields another ensnaring graph.

Figures

Figures reproduced from arXiv: 2502.06013 by the authors.

**Figure 1.** A beam of light refracts through a horizontal line. Combinatorial billiards is a new topic that merges ideas from dynamical algebraic combinatorics and mathematical billiards; it concerns billiard systems that are rigid and discretized in a way that allows them to be modeled combinatorially or algebraically [1, 3, 4, 10, 12, 13, 17]. Adams, Defant, and Striker [1] recently introduced combinatorial refraction billiar… view at source ↗

**Figure 2.** On the top is an unbounded combinatorial refraction billiard trajectory in Se3. On the bottom is the corresponding toric combinatorial refraction billiard trajectory, with each state represented both as an arrow in the torus and as a stone diagram [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

**Figure 3.** On the top is a bounded combinatorial refraction billiard trajectory in Se3. On the bottom is the corresponding toric combinatorial refraction billiard trajectory, with each state represented both as an arrow in the torus and as a stone diagram [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Four applications of the map ΘG. At each step, we represent a triple in Ξ5 as a stone diagram with the associated coin diagram drawn below. ΘG, we swap the replicas a and b. If a and b are adjacent in G, then the coin moves from a to b; otherwise, the coin stays put on…

**Figure 5.** Figure 5: A sequence transforming one pleasant stone diagram into another. 4. Complete Graphs and Cycles Our goal in this section is to exhibit two elementary families of ensnaring graphs: complete graphs and odd cycles. In what follows, we denote the n-vertex cycle graph by Cn;…

**Figure 6.** Figure 6: An illustration of the proof of Theorem 5.6 with n = 5. winding number with respect to O. Thus, all replicas of vertices in G2 have the same winding number − 1 n−1 (m1 + m) with respect to O. Let m2 = − 1 n−1 (m1 + m). Using the fact that the sum of all winding numbers…

**Figure 7.** Figure 7: A schematic illustration of the configuration described in Theorem 7.1. The edge between c and d is depicted as translucent to indicate that it may or may not be present. Theorem 7.1. Suppose a graph G has vertices a, b, c, and d with the following properties: (1) a ha…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 16 canonical work pages

[1]

Adams, C

A. Adams, C. Defant, and J. Striker. Toric promotion with reflections and refractions. Preprint: arXiv:2404.03649

work page arXiv
[2]

Baird-Smith, D

P. Baird-Smith, D. Davis, E. Fromm, and S. Iyer. Tiling billards on triangle tilings, and interval exchange transformations. J. Lond. Math. Soc., 109 (2024)

work page 2024
[3]

Barkley and R

G. Barkley and R. Liu. Channels, billiards, and perfect matching 2-divisibility. Electron. J. Combin., 28 (2021)

work page 2021
[4]

Barkley, C

G. Barkley, C. Defant, E. Hodges, N. Kravitz, and M. Lee. Bender–Knuth billiards in Coxeter groups. Forum Math. Sigma, 13 (2025)

work page 2025
[5]

V. L. Barutello, I. De Blasi, and S. Terracini. Chaotic dynamics in refraction galactic billiards. Nonlinearity, 36 (2023)

work page 2023
[6]

Davis, K

D. Davis, K. DiPietro, J. Rustad, and A. St. Laurent. Negative refraction and tiling billiards. Adv. Geom., 18 (2018), 133–159

work page 2018
[7]

Davis and W

D. Davis and W. P. Hooper. Periodicity and ergodicity in the trihexagonal tiling. Comment. Math. Helv., 93 (2018), 661–707

work page 2018
[8]

De Blasi and S

I. De Blasi and S. Terracini. Refraction periodic trajectories in central mass galaxies. Nonlinear Anal., 218 (2022)

work page 2022

Show all 17 references

[9]

De Blasi and S

I. De Blasi and S. Terracini. On some refraction billiards. Discrete Contin. Dyn. Syst., 43 (2023), 1269–1318

2023
[10]

C. Defant. Random combinatorial billiards and stoned exclusion processes. Preprint: arXiv:2406.07858

arXiv
[11]

C. Defant. Toric promotion. Proc. Amer. Math. Soc., 150 (2022)

2022
[12]

Defant and P

C. Defant and P. Jiradilok. Triangular-grid billiards and plabic graphs. Comb. Theory, 3 (2023)

2023
[13]

Defant, P

C. Defant, P. Jiradilok, and E. Mossel. Random subwords and billiard walks in affine Weyl groups. Preprint: arXiv:2501.11095

arXiv
[14]

M. Jay. Deviations for generalized tiling billiards in cyclic polygons. Preprint: arXiv:2402.15765

arXiv
[15]

Paris-Romaskevich

O. Paris-Romaskevich. Tiling billiards and Dynnikov’s helicoid. Trans. Moscow Math. Soc., 2021 (2021), 133– 147

2021
[16]

Smith, J

D. Smith, J. Pendry, and M. Wiltshire. Metamaterials and negative refractive index. Science, 305 (2004), 788– 792

2004
[17]

H. Zhu. The maximum number of cycles in a triangular-grid billiards system with a given perimeter. Preprint: arXiv:2309.00100. Department of Mathematics, Harvard University, Cambridge, MA 02138, USA Email address: colindefant@gmail.com Department of Mathematics, Massachusetts ...

arXiv

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.

[1] [1]

Adams, C

A. Adams, C. Defant, and J. Striker. Toric promotion with reflections and refractions. Preprint: arXiv:2404.03649

work page arXiv

[2] [2]

Baird-Smith, D

P. Baird-Smith, D. Davis, E. Fromm, and S. Iyer. Tiling billards on triangle tilings, and interval exchange transformations. J. Lond. Math. Soc., 109 (2024)

work page 2024

[3] [3]

Barkley and R

G. Barkley and R. Liu. Channels, billiards, and perfect matching 2-divisibility. Electron. J. Combin., 28 (2021)

work page 2021

[4] [4]

Barkley, C

G. Barkley, C. Defant, E. Hodges, N. Kravitz, and M. Lee. Bender–Knuth billiards in Coxeter groups. Forum Math. Sigma, 13 (2025)

work page 2025

[5] [5]

V. L. Barutello, I. De Blasi, and S. Terracini. Chaotic dynamics in refraction galactic billiards. Nonlinearity, 36 (2023)

work page 2023

[6] [6]

Davis, K

D. Davis, K. DiPietro, J. Rustad, and A. St. Laurent. Negative refraction and tiling billiards. Adv. Geom., 18 (2018), 133–159

work page 2018

[7] [7]

Davis and W

D. Davis and W. P. Hooper. Periodicity and ergodicity in the trihexagonal tiling. Comment. Math. Helv., 93 (2018), 661–707

work page 2018

[8] [8]

De Blasi and S

I. De Blasi and S. Terracini. Refraction periodic trajectories in central mass galaxies. Nonlinear Anal., 218 (2022)

work page 2022