\documentclass[11pt]{article} \usepackage[margin=1in]{geometry} \usepackage[utf8]{inputenc} \usepackage[T1]{fontenc} \usepackage{lmodern} \usepackage{amsmath,amssymb,amsthm} \usepackage{booktabs} \usepackage{tikz} \DeclareUnicodeCharacter{2200}{\ensuremath{\forall}} \DeclareUnicodeCharacter{2203}{\ensuremath{\exists}} \DeclareUnicodeCharacter{2192}{\ensuremath{\to}} \DeclareUnicodeCharacter{2260}{\ensuremath{\neq}} \DeclareUnicodeCharacter{2264}{\ensuremath{\le}} \DeclareUnicodeCharacter{2209}{\ensuremath{\notin}} \DeclareUnicodeCharacter{2227}{\ensuremath{\wedge}} \usepackage[title]{appendix} \usepackage[colorlinks=true,linkcolor=blue,citecolor=blue,urlcolor=blue,pdfauthor={P. Graham},pdftitle={Combinatorial periods of asymmetric billiard trajectories in the regular pentagon}]{hyperref} \newtheorem{theorem}{Theorem}[section] \newtheorem{proposition}[theorem]{Proposition} \newtheorem{corollary}[theorem]{Corollary} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{conjecture}[theorem]{Conjecture} \theoremstyle{definition} \newtheorem{definition}[theorem]{Definition} \theoremstyle{remark} \newtheorem{remark}[theorem]{Remark} \newtheorem{example}[theorem]{Example} \setlength{\emergencystretch}{3em} \newcommand{\Gp}{\Gamma^{+}} \title{Combinatorial periods of asymmetric billiard trajectories in the regular pentagon} \author{P. Graham\\ \small\url{https://orcid.org/0009-0007-8493-1505}} \date{November 10, 2026\\[2pt] \small Zenodo archive (Lean development in \texttt{lean}, \texttt{checks}, \texttt{data}): \url{https://doi.org/10.5281/zenodo.24119275}} \begin{document} \maketitle \begin{abstract} We prove that every even integer other than $2$, $12$, $14$ and $18$ is the combinatorial period, the number of bounces before repeating, of a periodic billiard trajectory in the regular pentagon that does not have five-fold rotational symmetry. Davis and Lelièvre conjectured this in 2018 and checked it up to period $4000$, and Kontorovich and Zhang proved it for almost every even number. As they explain, a theorem of Davis, Fuchs and Tabachnikov turns the question into one about vectors. Each vector $(a+b\varphi,c+d\varphi)$ obtained from $(0,1)$ by products of four explicit $2\times2$ matrices, where $\varphi$ is the golden ratio and $a,b,c,d$ are integers, has the half-period $\ell=a+b+c+d$, and when a congruence modulo $5$ holds it gives such a trajectory of period $n=2\ell$. Varying the power of one of the two triangular matrices in such a product gives half-periods in an arithmetic progression. We give $379$ of these progressions, which together contain every half-period $\ell\ge2$ except nine values. Six of the nine come from explicit trajectories checked in exact arithmetic; the other three are halves of excluded periods. The covering argument is checked in Lean, assuming the cited theorem. Whether the four excluded periods occur is not treated. This work has not yet been reviewed by experts. The proofs, the Lean formalization and the text were produced by AI models (from OpenAI and Anthropic), working under the author's direction. \end{abstract} \medskip \noindent\textit{Keywords:} polygonal billiards, regular pentagon, periodic trajectories, golden L, Hecke group, thin orbits. \noindent\textit{MSC 2020:} 37C83, 37E35, 11B25. \section{Introduction} \subsection{Main result} A billiard trajectory in a polygon moves in straight lines and reflects off the sides with equal angles of incidence and reflection; it is \emph{nonsingular} if it never hits a vertex. A nonsingular trajectory is \emph{periodic} if it returns to its initial point with its initial direction, and its \emph{combinatorial period} is the number of reflections in a least period, the ``number of edge crossings before repeating'' in the words of Davis and Lelièvre \cite[\S1]{DL}. Kontorovich and Zhang call a trajectory \emph{symmetric} when it has $5$-fold rotational symmetry \cite[\S1]{KZ}: the rotation $R$ by $2\pi/5$ about the center maps it to itself. Otherwise it is \emph{asymmetric}. Whether trajectories are taken with or without orientation does not matter here: if $R$ mapped a trajectory onto itself with its orientation reversed, then so would $R^5$, the identity, since $5$ is odd. \begin{theorem}\label{thm:main} For every even integer $n>0$ with $n\notin\{2,12,14,18\}$, there is a nonsingular periodic billiard trajectory in the regular pentagon whose combinatorial period is $n$ and which does not have $5$-fold rotational symmetry. \end{theorem} Whether the four excluded periods occur for asymmetric trajectories is a separate question, which Theorem~\ref{thm:main} does not address. \subsection{Background} Davis and Lelièvre showed that every positive multiple of $10$ is the combinatorial period of a rotationally symmetric trajectory \cite[Proposition~4.2]{DL}, and wrote \cite[\S4.1]{DL}: \begin{quote} Based on computer evidence, we conjecture the following: Conjecture 4.3. Every even number arises as the combinatorial period of an asymmetric periodic billiard trajectory on the pentagon, other than 2, 12, 14 and 18. We have verified this up to period 4000. \end{quote} Theorem~\ref{thm:main} is the existence statement of this conjecture. Kontorovich and Zhang restated it as their Conjecture~1.1 \cite[\S1]{KZ}. They proved \cite[Theorem~1.2]{KZ} that ``Conjecture~1.1 is true in density, that is, almost every even number arises'', with at most $\ll X/(\log X)^{1/40}$ exceptions up to $X$, and showed that evenness is the only congruence obstruction \cite[Theorem~1.4]{KZ}. They explained why the orbital circle method does not apply here \cite[Remark~1.5]{KZ}, and that their sieve method misses $\gg X/\log X$ values \cite[Remark~4.5]{KZ}. \subsection{Proof overview} Section~\ref{sec:cited} states the theorem of Davis, Fuchs and Tabachnikov \cite[Theorem~11]{DFT} that we use, in the form of \cite[Theorem~2.1]{KZ}. To each vector $v=(a+b\varphi,c+d\varphi)$ of an orbit $\Lambda$ of a semigroup $\Gp$ of $2\times2$ matrices it attaches the half-period $\ell(v)=a+b+c+d$, and when a congruence modulo $5$ holds it gives an asymmetric trajectory of period $n=2\ell(v)$. So Theorem~\ref{thm:main} asks for every half-period $\ell=n/2$ with $\ell\notin\{1,6,7,9\}$. Section~\ref{sec:families} shows that one-parameter families in $\Gp$ built on a parabolic generator give arithmetic progressions of half-periods, and Section~\ref{sec:covering} gives $379$ progressions that together contain every $\ell\ge2$ outside the set \[ E=\{4,5,6,7,9,15,21,29,51\}. \] Section~\ref{sec:small} treats $\ell\in\{4,5,15,21,29,51\}$, the periods $8$, $10$, $30$, $42$, $58$ and $102$, by six explicit trajectories, checked in exact arithmetic without the cited theorem. The remaining values $\ell\in\{1,6,7,9\}$ are the half-periods of the excluded periods. The new step is the covering: each family gives all large members of one residue class, so finitely many families whose classes exhaust the residues modulo a fixed number, with the few small values checked separately, give every half-period at once. The arithmetic of Sections~\ref{sec:families} and~\ref{sec:covering} is checked in Lean (Section~\ref{sec:verification}). \section{The cited theorem}\label{sec:cited} Let $\varphi=(1+\sqrt5)/2$, and let $\Gp$ be the semigroup generated by \[ A=\begin{pmatrix}1&\varphi\\0&1\end{pmatrix},\quad B=\begin{pmatrix}1&0\\\varphi&1\end{pmatrix},\quad C=\begin{pmatrix}\varphi&\varphi\\1&\varphi\end{pmatrix},\quad D=\begin{pmatrix}\varphi&1\\\varphi&\varphi\end{pmatrix}, \] as in \cite[\S2]{KZ}. All entries of the generators lie in $\mathbb{Z}_{\ge0}[\varphi]=\{p+q\varphi: p,q\in\mathbb{Z}_{\ge0}\}$, so every vector of the orbit $\Lambda=\Gp\cdot(0,1)^t$ has the form $v=(a+b\varphi,\,c+d\varphi)$ with $a,b,c,d\ge0$ integers, uniquely determined since $\varphi$ is irrational. Put \[ \ell(v)=a+b+c+d,\qquad \lambda(v)=(d-b)+2(c-a). \] Both are $\mathbb{Z}$-linear in $(a,b,c,d)$; we call $\ell(v)$ the \emph{half-period} of $v$. The geometry behind the cited theorem is as follows; the proof uses only its consequence, Corollary~\ref{cor:period}. Billiard trajectories in the pentagon correspond to straight lines on the double pentagon, a translation surface made of two copies of the pentagon, by the usual unfolding. Cutting and translating pieces of the double pentagon and applying a linear map, operations that preserve straight lines, gives another translation surface, the golden L \cite[\S2, Figure~2]{KZ}. A \emph{saddle connection} on such a surface is a straight segment that starts and ends at singular points, which come from the vertices of the pentagons, and passes through none in between; it is recorded by its displacement vector in the plane. A direction is \emph{periodic} when the straight lines in it are closed, apart from those that meet a singular point. In each periodic direction, Kontorovich and Zhang name one saddle connection short and one long, and all the proof needs is that every vector of $\Lambda$ is a long one. They write \cite[\S2]{KZ}: ``Each periodic direction in the golden L has a short and long saddle connection, and all long saddle connections with directions in the first quadrant are given by $\Lambda_L=\Gamma^+\cdot(0,1)^t$.'' They then state: \begin{theorem}[{\cite[Theorem~2.1]{KZ}, citing \cite[Theorem~11]{DFT}}]\label{thm:dft} ``Given a long saddle connection $(a+b\varphi,c+d\varphi)\in\Gamma^+\cdot(0,1)^t$ in the golden L, let $\ell=a+b+c+d$. The combinatorial period of the corresponding pentagon billiard trajectory is given either by $L=2\ell$ or $L=10\ell$, depending on whether $(d-b)+2(c-a)\equiv0\pmod5$ or not (resp.), and corresponding to whether the associated trajectory is asymmetric or not.'' \end{theorem} By the sentence before it, the hypothesis of Theorem~\ref{thm:dft} holds for every $v\in\Lambda$. We use only its first case. \begin{corollary}\label{cor:period} If $v\in\Lambda$ and $\lambda(v)\equiv0\pmod5$, then $2\ell(v)$ is the combinatorial period of an asymmetric periodic billiard trajectory in the regular pentagon. \end{corollary} Theorem~\ref{thm:main} therefore follows from Proposition~\ref{prop:cover} below, on values of $\ell$, together with Proposition~\ref{prop:small} for six small periods. \section{Parabolic families}\label{sec:families} Write $v\in\Lambda$ as the column $(a,b,c,d)\in\mathbb{Z}_{\ge0}^4$. Using $\varphi^2=\varphi+1$, each generator acts on these coordinates by a nonnegative integer matrix; for example $A$ sends $(a,b,c,d)$ to $(a+d,\,b+c+d,\,c,\,d)$. Two of the generators are parabolic: \[ A=I+N_A,\quad N_A=\begin{pmatrix}0&\varphi\\0&0\end{pmatrix},\qquad B=I+N_B,\quad N_B=\begin{pmatrix}0&0\\\varphi&0\end{pmatrix}, \qquad N_A^2=N_B^2=0, \] so $P^n=I+nN_P$ for $P\in\{A,B\}$ and all $n\ge0$. \begin{lemma}\label{lem:family} Let $u,w$ be words in $A,B,C,D$, let $P\in\{A,B\}$, and put $v_n=uP^nw\cdot(0,1)^t\in\Lambda$. Then $\ell(v_n)=\ell_0+n\alpha$ and $\lambda(v_n)=\lambda_0+n\beta$ for integers $\ell_0,\alpha,\lambda_0,\beta$, where $\alpha=\ell(uN_Pw\cdot(0,1)^t)\ge0$. If $\alpha>0$ and $\beta\not\equiv0\pmod5$, let $n_0\in\{0,\dots,4\}$ satisfy $\lambda_0+n_0\beta\equiv0\pmod5$; then every $s+jm$ with $j\ge0$, where $s=\ell_0+n_0\alpha$ and $m=5\alpha$, is $\ell(v)$ for some $v\in\Lambda$ with $\lambda(v)\equiv0\pmod5$. If $\alpha>0$, $\beta\equiv0$ and $\lambda_0+n_0\beta\equiv0\pmod5$ for some $n_0\ge0$, the same holds with $s=\ell_0+n_0\alpha$ and $m=\alpha$. \end{lemma} \begin{proof} Since $u$ acts linearly, $v_n=uw\cdot(0,1)^t+n\,uN_Pw\cdot(0,1)^t$, and $\ell$ and $\lambda$ are linear. The orbit vectors $v_{n_0+5j}$, or $v_{n_0+j}$ when $\beta\equiv0$, have $\lambda\equiv0\pmod5$ and $\ell=s+jm$. \end{proof} \begin{example} Take $u$ and $w$ empty and $P=A$. Then $A^n(0,1)^t=(n\varphi,1)^t$, so $(a,b,c,d)=(0,n,1,0)$, $\ell=n+1$ and $\lambda=2-n$. With $n=2+5j$ we get asymmetric vectors with $\ell=3+5j$, so every $2\ell$ with $\ell\equiv3\pmod5$ and $\ell\ge3$ is an asymmetric period. \end{example} A single family covers one residue class modulo its $m$, from its start $s$ on. Kontorovich and Zhang's proof of their density theorem also produces periods from products of parabolic powers, the vectors obtained from $B^mA^nB^k$ \cite[(4.1)]{KZ}, with exponents restricted to progressions modulo $5$ so that the trajectories stay asymmetric \cite[\S4]{KZ}; there $\ell$ is cubic in the exponents, and a sieve gives almost all values. Here only one exponent varies, so $\ell$ is linear in it, and the question becomes one of covering residue classes. \section{The covering}\label{sec:covering} \begin{proposition}\label{prop:cover} For every integer $\ell\ge2$ with $\ell\notin E$, there is $v\in\Lambda$ with $\ell(v)=\ell$ and $\lambda(v)\equiv0\pmod5$. \end{proposition} The proof is a finite verification organized as a tree. Let \[ M=154\,954\,800=2^4\cdot3^2\cdot5^2\cdot7\cdot11\cdot13\cdot43, \] and fix the order of its prime factors $(p_1,\dots,p_{12})=(5,5,2,3,2,7,2,3,2,11,13,43)$. A \emph{node} is a pair $(Q,r)$ with $Q=p_1\cdots p_k$ for some $0\le k\le12$ and $0\le r0$ be even with $n\notin\{2,12,14,18\}$, and let $\ell=n/2$, so $\ell\ge1$ and $\ell\notin\{1,6,7,9\}$. If $\ell\in\{4,5,15,21,29,51\}$, Proposition~\ref{prop:small} gives the trajectory. Otherwise $\ell\ge2$ and $\ell\notin E$, and Proposition~\ref{prop:cover} gives $v\in\Lambda$ with $\ell(v)=\ell$ and $\lambda(v)\equiv0\pmod5$; by Corollary~\ref{cor:period}, $n=2\ell(v)$ is the combinatorial period of an asymmetric periodic trajectory. \end{proof} \section{Verification}\label{sec:verification} The deduction of Theorem~\ref{thm:main} from Theorem~\ref{thm:dft} and Proposition~\ref{prop:small} is checked in Lean~4 \cite{Lean}, without external libraries. The Lean statement defines $\mathbb{Z}_{\ge0}[\varphi]$, the generators, the orbit $\Lambda$ as an inductive predicate, $\ell$ and the condition $\lambda\equiv0\pmod5$. Billiards are not formalized: ``some asymmetric periodic trajectory has combinatorial period $n$'' is an opaque predicate $\mathrm{AsymPeriod}(n)$, which enters only through Theorem~\ref{thm:dft}, stated as the axiom \[ \forall v,\ v\in\Lambda\ \to\ \lambda(v)\equiv0\pmod5\ \to\ \mathrm{AsymPeriod}(2\ell(v)), \] and through the hypothesis that the six periods of Proposition~\ref{prop:small} occur. The main formal theorem, \texttt{dl\_conjecture\_4\_3\_existence}, says: if $\mathrm{AsymPeriod}(n)$ holds for $n=8,10,30,42,58,102$, then it holds for every even $n>0$ other than $2,12,14,18$. With the axiom read as Corollary~\ref{cor:period} and the hypothesis as Proposition~\ref{prop:small}, this is Theorem~\ref{thm:main}. Its proof uses the axiom and Lean's standard axioms only. Proposition~\ref{prop:cover} is the formal theorem \texttt{long\_half\_periods}, proved with no axiom beyond Lean's standard ones: Lemma~\ref{lem:family} and Lemma~\ref{lem:leaf} are proved in general, and the $379$ families and the tree of Section~\ref{sec:covering} are checked by the kernel's evaluation of a decision procedure. Proposition~\ref{prop:small} is checked outside Lean by the two exact programs, which also reject altered certificates. Appendix~\ref{app:lean} gives the formal statements, and the accompanying README gives the commands to reproduce both checks. \section*{Acknowledgments} This paper builds on the work of Diana Davis and Samuel Lelièvre, who posed the question and computed the periods up to $4000$, and of Alex Kontorovich and Xin Zhang, whose reformulation in terms of $\Gp$ and statement of the period theorem are the starting point of the proof. The period theorem is due to Diana Davis, Dmitry Fuchs and Serge Tabachnikov. \begin{thebibliography}{9} \bibitem{DFT} D.~Davis, D.~Fuchs and S.~Tabachnikov, \emph{Periodic trajectories in the regular pentagon}, Mosc. Math. J. \textbf{11} (2011), no.~3, 439--461; \href{https://arxiv.org/abs/1102.1005}{arXiv:1102.1005}. \bibitem{DL} D.~Davis and S.~Lelièvre, \emph{Periodic paths on the pentagon, double pentagon and golden L}, \href{https://arxiv.org/abs/1810.11310v2}{arXiv:1810.11310v2} (2024). \bibitem{KZ} A.~Kontorovich and X.~Zhang, \emph{On the local-global conjecture for combinatorial period lengths of closed billiards on the regular pentagon}, Int. Math. Res. Not. IMRN \textbf{2025} (2025), no.~15, rnaf227, \href{https://doi.org/10.1093/imrn/rnaf227}{doi:10.1093/imrn/rnaf227}; \href{https://arxiv.org/abs/2409.10682}{arXiv:2409.10682}. \bibitem{Lean} L.~de~Moura and S.~Ullrich, \emph{The Lean 4 theorem prover and programming language}, in: Automated Deduction -- CADE 28, Lecture Notes in Comput. Sci. \textbf{12699}, Springer, 2021, 625--635. \end{thebibliography} \begin{appendices} \section{The formal statements}\label{app:lean} The definitions are in \texttt{Pentagon/Statement.lean}. An element $p+q\varphi$ of $\mathbb{Z}_{\ge0}[\varphi]$ is a pair of natural numbers with the product $(p+q\varphi)(r+s\varphi)=pr+qs+(ps+qr+qs)\varphi$; a vector is a pair of these and a matrix a quadruple. The generators are \texttt{A}, \texttt{B}, \texttt{C}, \texttt{D}, written entrywise as in Section~\ref{sec:cited}. \texttt{Orbit v} is the least predicate containing $(0,1)$ and closed under the four generators; \texttt{ell v} is $a+b+c+d$ and \texttt{Asym v} is $((d-b)+2(c-a))\bmod 5=0$ in the integers. The cited theorem is the axiom \begin{verbatim} axiom kz_thm_2_1 : ∀ v : Vec, Orbit v → Asym v → AsymPeriod (2 * ell v) \end{verbatim} whose documentation quotes Theorem~\ref{thm:dft} and the sentence before it. The main theorem, in \texttt{Pentagon/Proof.lean}, is \begin{verbatim} theorem dl_conjecture_4_3_existence (small : AsymPeriod 8 ∧ AsymPeriod 10 ∧ AsymPeriod 30 ∧ AsymPeriod 42 ∧ AsymPeriod 58 ∧ AsymPeriod 102) : ∀ n : Nat, 0 < n → n % 2 = 0 → n ≠ 2 → n ≠ 12 → n ≠ 14 → n ≠ 18 → AsymPeriod n \end{verbatim} and Proposition~\ref{prop:cover} is \begin{verbatim} theorem long_half_periods (l : Nat) (h2 : 2 ≤ l) (hl : l ∉ exceptions) : ∃ v, Orbit v ∧ ell v = l ∧ Asym v \end{verbatim} with \texttt{exceptions} the list $4,5,6,7,9,15,21,29,51$. The file \texttt{Pentagon/Cert.lean} lists the $379$ families by the depth of the nodes they close, each as $(m,s,w,P,u,n_0)$, and the prime order. \texttt{famOK} checks a family against Lemma~\ref{lem:family} by computing $uw\cdot(0,1)^t$ and $uN_Pw\cdot(0,1)^t$, and \texttt{chk} walks the tree, expanding a node until \texttt{leafOK}, the closing condition, holds; both are evaluated by the kernel (\texttt{decide +kernel}), and their soundness lemmas \texttt{fam\_sound}, \texttt{leaf\_sound} and \texttt{chk\_sound} are proved. The build file \texttt{PublicChecks.lean} restates these theorems with their full types and checks their axioms: \texttt{long\_half\_periods} uses only the standard axioms of Lean's logic, and \texttt{dl\_conjecture\_4\_3\_existence} uses these and the cited axiom. The development also proves \texttt{exceptions\_combinatorial}: no $v\in\Lambda$ with $\lambda(v)\equiv0\pmod5$ has $\ell(v)$ or $\ell(\varphi v)$ in $\{1,6,7,9\}$, a statement about the orbit that Theorem~\ref{thm:main} does not use. \end{appendices} \end{document}