OctoInformBeta

OpenAI preprint, September 2026 · §2. The double and angular coefficients

§2 (setting): the oriented double and the regular two-sided flow: Vertex exceptions, complete trajectories and measure-preserving time maps

Regular billiards and their measured double.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

A billiard state specifies a position in the plane and a unit direction of motion.✦ Let QQ be a triangle with noncollinear vertices qi∈ℂq_i\in\mathbb C, indexed cyclically by i∈ℤ/3ℤi\in\mathbb Z/3\mathbb Z. The open table TT, open sides SiS_i and side tangent vectors eie_i of the triangle QQ are, respectively, the planar region, line segments and vectors defined by (1). T=int⁡conv⁡{q0,q1,q2},Si=(qi,qi+1),ei=qi+1−qi.T=\operatorname{int}\operatorname{conv}\{q_0,q_1,q_2\},\qquad S_i=(q_i,q_{i+1}),\qquad e_i=q_{i+1}-q_i.(1) The interval in the definition of SiS_i denotes an open line segment. Specular reflection in the direction of a nonzero vector e∈ℂe\in\mathbb C is the real-linear map ReR_e in (2); the affine reflection ρi\rho_i in side ii is also specified there. Re(v)=ee¯v¯,ρi(x)=qi+Rei(x−qi).R_e(v)=\frac{e}{\overline e}\,\overline v,\qquad \rho_i(x)=q_i+R_{e_i}(x-q_i).(2)

The phase space is 𝒫=ℂ×S1\mathcal P=\mathbb C\times S^1, where S1S^1 is the circle of unit complex numbers.✦ The uniform angular probability η\eta on S1S^1 is the image of dθ/(2π)d\theta/(2\pi) on [0,2π)[0,2\pi) under θ↦eiθ\theta\mapsto e^{i\theta}.✦ The normalized area-angular measure μ\mu on 𝒫\mathcal P is defined by (3), where dAdA is planar area. μ=dA|Tarea⁡(T)⊗η.\mu=\frac{dA|_T}{\operatorname{area}(T)}\otimes\eta.(3)✦

For v∈ℂv\in\mathbb C, let Lv(x)=Im⁡(xv¯)L_v(x)=\operatorname{Im}(x\overline v) be the oriented angular momentum functional.✦ The set ApA_p of states aiming at a point p∈ℂp\in\mathbb C is defined by (4). Ap={(x,v)∈𝒫:Lv(x−p)=0}.A_p=\{(x,v)\in\mathcal P:L_v(x-p)=0\}.(4)✦ For a finite word w=(j1,…,jm)w=(j_1,\ldots,j_m) of side indices, the unfolded vertex qw,iq_{w,i} is the successive reflected image in (5), with the empty word giving qiq_i. The vertex exceptional set EQE_Q in 𝒫\mathcal P is the union of the aiming sets Aqw,iA_{q_{w,i}} over all finite words ww of side indices and all vertex indices ii, as in (5). qw,i=ρj1∘⋯∘ρjm(qi),EQ=⋃w⋃i∈ℤ/3ℤAqw,i.q_{w,i}=\rho_{j_1}\circ\cdots\circ\rho_{j_m}(q_i),\qquad E_Q=\bigcup_w\ \bigcup_{i\in\mathbb Z/3\mathbb Z} A_{q_{w,i}}.(5)

Complete Flight Chains

A complete regular flight chain cc through a state z=(x,v)∈𝒫z=(x,v)\in\mathcal P is a sequence of collision times τn∈ℝ\tau_n\in\mathbb R, collision points pn∈ℂp_n\in\mathbb C, outgoing directions vn∈S1v_n\in S^1 and side indices jn∈ℤ/3ℤj_n\in\mathbb Z/3\mathbb Z, indexed by n∈ℤn\in\mathbb Z. Its times are strictly increasing and unbounded in both directions, with τ0<0<τ1\tau_0<0<\tau_1. The collision, flight and reflection conditions are (6). pn∈Sjn,pn+1=pn+(τn+1−τn)vn,vn=Rejn(vn−1),x=p0−τ0v0,v=v0.\begin{aligned} p_n&\in S_{j_n},& p_{n+1}&=p_n+(\tau_{n+1}-\tau_n)v_n,\\ v_n&=R_{e_{j_n}}(v_{n-1}),& x&=p_0-\tau_0v_0,\qquad v=v_0. \end{aligned}(6) The interior condition is (7). pn+(t−τn)vn∈T(τn<t<τn+1,n∈ℤ).p_n+(t-\tau_n)v_n\in T \qquad(\tau_n<t<\tau_{n+1},\ n\in\mathbb Z).(7) Write ℱQ(z)\mathcal F_Q(z) for the set of these chains. For c∈ℱQ(z)c\in\mathcal F_Q(z), the state c(t)∈𝒫c(t)\in\mathcal P is its position and direction at time tt, using the outgoing direction at a collision, and nc(t)∈ℤn_c(t)\in\mathbb Z is the index of the flight containing that time. For a state z∈𝒫z\in\mathcal P admitting a complete regular chain cc, the billiard time map Φt:𝒫→𝒫\Phi_t:\mathcal P\to\mathcal P is given by Φt(z)=c(t)\Phi_t(z)=c(t); it fixes zz when no such chain exists. For a state zz on a complete regular trajectory, its collision index Nt(z)∈ℤN_t(z)\in\mathbb Z is the label of the flight reached at time tt, with the initial flight labeled 00.

Valid collision data cc for a phase state z∈𝒫z\in\mathcal P are sequences of collision times, collision points, outgoing directions and side indices whose times are strictly increasing and unbounded in both directions, with τ0<0<τ1\tau_0<0<\tau_1, and which satisfy (6) and the midpoint instance of (7) on every flight. Let 𝒞Q\mathcal C_Q be the space of pairs (z,c)(z,c) with valid collision data cc for zz✦, and let I:𝒞Q→𝒫I:\mathcal C_Q\to\mathcal P, I(z,c)=zI(z,c)=z, be its initial-state map.✦ The trajectory recovery map J:𝒫→𝒞QJ:\mathcal P\to\mathcal C_Q extends the inverse of the initial-state map II from I(𝒞Q)I(\mathcal C_Q) to phase space, so J(I(c))=cJ(I(c))=c for every c∈𝒞Qc\in\mathcal C_Q.

The Claim

Definition 1 (Measured double and regular flow). Fix a nondegenerate triangle QQ. The double phase space 𝒟\mathcal D is the product in (8). 𝒟=𝒫×ℤ/2ℤ.\mathcal D=\mathcal P\times\mathbb Z/2\mathbb Z.(8)✦ The uniform sheet probability σ\sigma on ℤ/2ℤ\mathbb Z/2\mathbb Z is the probability measure defined from counting measure by (9). σ=12count⁡ℤ/2ℤ.\sigma=\frac12\,\operatorname{count}_{\mathbb Z/2\mathbb Z}.(9)✦ The double Liouville probability ν\nu on the double phase space 𝒟\mathcal D is the product measure in (10). ν=μ⊗σ.\nu=\mu\otimes\sigma.(10) The flow Ψt\Psi_t on the double is defined by (11), where [Nt(z)]2[N_t(z)]_2 is the residue class of the collision index modulo two. Ψt(z,a)=(Φt(z),[Nt(z)]2+a)(z∈𝒫,a∈ℤ/2ℤ).\Psi_t(z,a)=\bigl(\Phi_t(z),[N_t(z)]_2+a\bigr) \qquad(z\in\mathcal P,\ a\in\mathbb Z/2\mathbb Z).(11)✦

The vertex exceptional set has the measure specified in (12). μ(EQ)=0.\mu(E_Q)=0.(12)✦ For z=(x,v)z=(x,v) with x∈Tx\in T and z∉EQz\notin E_Q, complete chains exist as in (13). ℱQ(z)≠⌀.\mathcal F_Q(z)\ne\varnothing.(13)✦ Almost-everywhere existence is recorded in (14). ℱQ(z)≠⌀for μ-almost every z∈𝒫.\mathcal F_Q(z)\ne\varnothing\qquad\text{for $\mu$-almost every }z\in\mathcal P.(14)✦ For any phase state zz and any chains c,d∈ℱQ(z)c,d\in\mathcal F_Q(z), uniqueness takes the form (15). c=d.c=d.(15)✦

For every t∈ℝt\in\mathbb R, the map Φt\Phi_t is measurable and satisfies (16). (Φt)*μ=μ.(\Phi_t)_*\mu=\mu.(16)✦ The billiard time maps satisfy the flow law (17) for all s,t∈ℝs,t\in\mathbb R. Φs+t(z)=Φs(Φt(z))for μ-almost every z∈𝒫.\Phi_{s+t}(z)=\Phi_s(\Phi_t(z)) \qquad\text{for $\mu$-almost every }z\in\mathcal P.(17)✦ The time maps Ψt\Psi_t on the double are measurable and satisfy (18) for all t∈ℝt\in\mathbb R. (Ψt)*ν=ν.(\Psi_t)_*\nu=\nu.(18)✦ On the double, the flow law is (19) for all s,t∈ℝs,t\in\mathbb R. Ψs+t(ζ)=Ψs(Ψt(ζ))for ν-almost every ζ∈𝒟.\Psi_{s+t}(\zeta)=\Psi_s(\Psi_t(\zeta)) \qquad\text{for $\nu$-almost every }\zeta\in\mathcal D.(19)✦

In the paper, the result reads: gluing two oppositely oriented copies of the closed triangle along matching sides and removing their vertices gives a connected flat surface MM with cone angles 2α,2β,2γ2\alpha,2\beta,2\gamma in its completion, where α,β,γ\alpha,\beta,\gamma are the triangle’s angles; its unit tangent bundle has normalized Liouville measure ν\nu, folding sends the geodesic flow to the billiard flow and ν\nu to μ\mu, vertex-hit states are null, no other finite-time failure occurs, and regular time maps preserve measure.

What Is Checked

The statements are over nondegenerate triangles QQ, complete two-sided specular flight chains in 𝒫\mathcal P, and the measured double (𝒟,ν)(\mathcal D,\nu). Valid collision data for zz determine a chain in ℱQ(z)\mathcal F_Q(z) with the same collision sequences.✦ The initial-state map I:𝒞Q→𝒫I:\mathcal C_Q\to\mathcal P is a measurable embedding.✦