OctoInformBeta

OpenAI preprint, September 2026 · §5. Uniform control of returning geodesics

Proposition 5.5 [All neck passages]: Both endpoint frames control finite segments with boundary or vanishing-field endpoints and compactly supported variations on an open interval

Intrinsic index across all neck passages.

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

The geodesics may have arbitrary affine parametrization.

Let ℍ\mathbb H be the upper half-plane and let Γ\Gamma be the modular subgroup with the generators in (1); its effective action is its image in the permutations of ℍ\mathbb H. Let 𝕊=ℝ/6ℤ\mathbb S=\mathbb R/6\mathbb Z be the flat quotient circle✦ and let Σ=ℍ/Γ\Sigma=\mathbb H/\Gamma be the quotient hyperbolic surface.✦ ϕ=(1112),ψ=(1−1−12),Γ=⟨ϕ,ψ,−I⟩⊂SL2(ℤ).\phi=\begin{pmatrix}1&1\\1&2\end{pmatrix},\qquad \psi=\begin{pmatrix}1&-1\\-1&2\end{pmatrix},\qquad \Gamma=\langle\phi,\psi,-I\rangle\subset\mathrm{SL}_2(\mathbb Z).(1) The surface height hh on Σ\Sigma is the real-valued function given by the logarithm of the maximal imaginary part of a modular translate of a representative.✦ The open truncated surface Σ<log⁡8={q:h(q)<log⁡8}\Sigma_{<\log8}=\{q:h(q)<\log8\} is the subset of Σ\Sigma below height log⁡8\log8.✦ For a threshold rr, the surface cusp is the image of {z:Im⁡z>r}\{z:\operatorname{Im}z>r\} under the surface quotient projection.✦

For a real period bb, the circle deck group bℤb\mathbb Z is the group of translations of ℝ\mathbb R by integer multiples of bb.✦

Put X=Σ<log⁡8×𝕊X=\Sigma_{<\log8}\times\mathbb S.✦ The torus collar O⊂XO\subset X is the subset consisting of points whose surface coordinate is in the cusp above 22.✦ Its collar coordinates identify it with (2,8)×𝕊2(2,8)\times\mathbb S^2, using the surface cusp coordinates and the original circle coordinate.✦ The torus automorphism T(x,y)=(x+y,x)T(x,y)=(x+y,x) has inverse T−1(x,y)=(y,x−y)T^{-1}(x,y)=(y,x-y).✦ The glued three-manifold MM is obtained from two copies of XX by identifying their collars through (r,x,y)↦(16/r,T−1(x,y))(r,x,y)\mapsto(16/r,T^{-1}(x,y)).✦ Write ι−\iota_- and ι+\iota_+ for the copy inclusions. The open neck N=ι−(O)N=\iota_-(O) is a subset of MM.✦ The radial inversion of (2,8)(2,8) is the self-inverse homeomorphism r↦16/rr\mapsto16/r.✦

The Metric and Its Cometrics

Let tt denote neck height, and write θ\theta for the standard smooth transition from zero to one. The one-sided cutoff aa, its integral AA, and the shear cutoff χ\chi are the functions in (2). a(t)={exp⁡(−1/(2/3−t)),t<2/3,0,t≥2/3,A(t)=∫0ta(u)du,χ(t)=θ(6(t−5/12)).\begin{aligned} a(t)&=\begin{cases} \exp\!\bigl(-1/(2/3-t)\bigr),&t<2/3,\\ 0,&t\geq2/3, \end{cases} &A(t)&=\int_0^t a(u)\,du,\\ \chi(t)&=\theta\!\left(6(t-5/12)\right). \end{aligned}(2) For the neck parameter δ\delta, the diagonal coefficients d1,d2d_1,d_2, the cometric shear σ\sigma, and the real 2×22\times2 cometric PP are given by (3). Transverse vectors belong to ℝ2\mathbb R^2, with squared norm |q|2=q12+q22|q|^2=q_1^2+q_2^2; the quadratic form of a matrix BB is q𝖳Bqq^{\mathsf T}Bq. d1(δ,t)=1+δA(t),d2(δ,t)=1+δA(1−t),σ(δ,t)=δ2χ(t),Pδ(t)=(d1+d2σ2d2σd2σd2).\begin{aligned} d_1(\delta,t)&=1+\delta A(t),& d_2(\delta,t)&=1+\delta A(1-t),& \sigma(\delta,t)&=\delta^2\chi(t),\\ P_\delta(t)&= \begin{pmatrix} d_1+d_2\sigma^2&d_2\sigma\\ d_2\sigma&d_2 \end{pmatrix}. \end{aligned}(3) In the matrix in (3), all coefficients are evaluated at (δ,t)(\delta,t).

The collar denominator is d1(δ,t)d_1(\delta,t).✦ Its terminal decay rate is δa(t)/(2d1(δ,t))\delta a(t)/(2d_1(\delta,t)).✦ The collar transition λ\lambda, logarithmic decay kk, and warping function ℓ\ell are defined in (4). λ(t)=θ(2t+3/2),k(δ,t)=1−λ(t)+λ(t)δa(t)2d1(δ,t),ℓ(δ,t)=exp⁡(−∫0tk(δ,u)du).\begin{aligned} \lambda(t)&=\theta(2t+3/2),\\ k(\delta,t)&=1-\lambda(t) +\lambda(t)\frac{\delta a(t)}{2d_1(\delta,t)},\\ \ell(\delta,t)&=\exp\!\left(-\int_0^t k(\delta,u)\,du\right). \end{aligned}(4) For 0<δ≤1/20<\delta\leq1/2 and a positive period BB, the interpolating neck metric is the quotient cylinder metric in (5) on ℝ×𝕊2\mathbb R\times\mathbb S^2.✦ With the scale functions r1r_1 and r2r_2 given in (5), its lift to ℝ×ℝ2\mathbb R\times\mathbb R^2 is the Euclidean metric pulled back by the triangular moving frame (v0,v1,v2)↦(v0,r1(v1−χv2),r2v2)(v_0,v_1,v_2)\mapsto(v_0,r_1(v_1-\chi v_2),r_2v_2).✦ r1(δ,B,t)=δ2B6ℓ(δ,t),r2(δ,B,t)=B6ℓ(δ,1−t),gneck,δ,B=dt2+r1(δ,B,t)2(dx−χ(t)dy)2+r2(δ,B,t)2dy2.\begin{aligned} r_1(\delta,B,t)&=\frac{\delta^2B}{6}\ell(\delta,t),& r_2(\delta,B,t)&=\frac{B}{6}\ell(\delta,1-t),\\ g_{\mathrm{neck},\delta,B} &=dt^2+r_1(\delta,B,t)^2(dx-\chi(t)\,dy)^2 +r_2(\delta,B,t)^2dy^2. \end{aligned}(5) The cylinder deck group is 6ℤ×6ℤ6\mathbb Z\times6\mathbb Z, acting by translations in the two horizontal coordinates and fixing height.✦

For a length parameter Λ\Lambda, the radial cylinder map FΛF_\Lambda changes height by the profile RΛR_\Lambda in (6) and leaves the torus coordinates unchanged.✦ The radial slope is RΛ′(t)R_\Lambda'(t), and its lift makes the same change on ℝ×ℝ2\mathbb R\times\mathbb R^2. η(t)=θ(4t−1),RΛ(t)=(1−η(t))(t−Λ−1)+η(t)t−log⁡2log⁡4.\eta(t)=\theta(4t-1),\qquad R_\Lambda(t)=(1-\eta(t))(t-\Lambda-1) +\eta(t)\frac{t-\log2}{\log4}.(6) The reversing cylinder map is H(t,x,y)=(1−t,T(x,y))H(t,x,y)=(1-t,T(x,y)).✦ Set the matching period BδB_\delta, left length Λ−\Lambda_-, and circle scales c−,c+c_-,c_+ as in (7). Bδ=6e4ℓ(δ,−1),Λ−=4−2log⁡δ,c−=Bδ6ℓ(δ,1),c+=δ2Bδ6ℓ(δ,1).\begin{aligned} B_\delta&=\frac{6}{e^4\ell(\delta,-1)},& \Lambda_-&=4-2\log\delta,\\ c_-&=\frac{B_\delta}{6}\ell(\delta,1),& c_+&=\frac{\delta^2B_\delta}{6}\ell(\delta,1). \end{aligned}(7) The left cylindrical metric is FΛ−*gneck,δ,BδF_{\Lambda_-}^*g_{\mathrm{neck},\delta,B_\delta}; the right cylindrical metric is F4*H*gneck,δ,BδF_4^*H^*g_{\mathrm{neck},\delta,B_\delta}. The radial transition function η\eta is the standard smooth transition evaluated at 4t−14t-1, as in (6). The height reflection is the self-inverse homeomorphism of ℝ\mathbb R sending tt to 1−t1-t. For a nonzero real number aa, the radial linear dilation of ℝ×ℝ2\mathbb R\times\mathbb R^2 is the linear automorphism (v0,v1,v2)↦(av0,v1,v2)(v_0,v_1,v_2)\mapsto(av_0,v_1,v_2); its inverse divides the height component by aa. For Λ≥4\Lambda\geq4, the radial pullback of a cylinder metric gg is FΛ*gF_\Lambda^*g. The switched cylinder metric associated with a cylinder metric gg is its pullback H*gH^*g under the reversing cylinder map.

The hyperbolic surface metric gΣg_\Sigma is the quotient of (dx2+dy2)/y2(dx^2+dy^2)/y^2 on ℍ\mathbb H.✦ On Σ×𝕊\Sigma\times\mathbb S, the core product metric with circle scale c≠0c\ne0 is gΣ+c2dϑ2g_\Sigma+c^2d\vartheta^2, where dϑ2d\vartheta^2 is the flat circle metric.✦ The core domain of the core product metric on Σ×𝕊\Sigma\times\mathbb S is the region h<1/4h<1/4.✦ Above the cusp threshold e1/8e^{1/8}, product cusp coordinates give (r,x,y)(r,x,y) with r>e1/8r>e^{1/8}; the logarithmic cylinder coordinates send this point to (log⁡r,x,y)(\log r,x,y), with inverse given by exponentiating the height.✦ A cylinder metric is pulled back to this product cusp. The cusp matching condition requires the lifted cylinder metric below height 1/41/4 to have the form (8). dt2+e−2tdx2+c2dy2.dt^2+e^{-2t}dx^2+c^2dy^2.(8) The complete left and right piece metrics are the metrics obtained by gluing these core and cusp metrics with scales c−c_- and c+c_+, respectively. The truncated left and right piece metrics on XX are the restrictions of the corresponding complete piece metrics. The piece charts identify the two copies of XX with their images in MM✦; gluing the truncated piece metrics in the inverse piece charts defines the compact metric on MM.✦ Denote by gδg_\delta this smooth Riemannian metric expressed in real three-dimensional coordinates.✦ A local Riemannian metric on a manifold is a smoothly varying Riemannian inner product on its tangent spaces over an open domain.✦

Passage Bounds and Smallness

The endpoint cometrics P−P_- and P+P_+ are the matrix families in (9), where τ\tau is inward endpoint height and C2C_2 is the endpoint frame matrix. C2(δ)=(011−δ2),P−(τ)=Pδ(τ),P+(τ)=C2(δ)Pδ(1−τ)C2(δ)𝖳.C_2(\delta)=\begin{pmatrix}0&1\\1&-\delta^2\end{pmatrix},\qquad P_-(\tau)=P_\delta(\tau),\qquad P_+(\tau)=C_2(\delta)P_\delta(1-\tau)C_2(\delta)^{\mathsf T}.(9) Passage cometric bounds with constant CC are the estimates (10), required for both endpoint frames, every τ∈[0,1]\tau\in[0,1], and every q∈ℝ2q\in\mathbb R^2; primes differentiate inward height. |q|29≤q𝖳P±(τ)q≤18|q|2,|P±′(τ)q|2≤C2δ2|q|2,0≤q𝖳(−P±″(τ))q,|P±′(τ)q|2≤Cδq𝖳(−P±″(τ))q.\begin{aligned} \frac{|q|^2}{9}&\leq q^{\mathsf T}P_\pm(\tau)q\leq18|q|^2,\\ |P_\pm'(\tau)q|^2&\leq C^2\delta^2|q|^2,\\ 0&\leq q^{\mathsf T}(-P_\pm''(\tau))q,\\ |P_\pm'(\tau)q|^2 &\leq C\delta\,q^{\mathsf T}(-P_\pm''(\tau))q. \end{aligned}(10)✦

The turning constants a0,βa_0,\beta and the mixed coefficient bound Cm(C)C_m(C) are defined in (11). a0=e−3/2,β=1+e−9/262,Cm(C)=a(0)+2Ca(0)+22Ca(0)/3.\begin{aligned} a_0&=e^{-3/2},& \beta&=\frac{1+e^{-9/26}}2,\\ C_m(C)&=a(0)+\sqrt{2Ca(0)}+2\sqrt{2Ca(0)/3}. \end{aligned}(11) For real coefficients Ce,CmC_e,C_m, the smallness condition on δ\delta is (12). 0<δ≤1,δ≤12Cm,δ≤1−β2Cm+Cea0.0<\delta\leq1,\qquad \delta\leq\frac1{2C_m},\qquad \delta\leq\frac{1-\beta}{2C_m+C_ea_0}.(12)✦

The Intrinsic Index

Let gg be a smooth metric on MM. In a chart cc, let (gij)(g_{ij}) be its metric matrix in the standard coordinate basis eie_i, and let ∂i\partial_i denote differentiation in direction eie_i. Write Γjki\Gamma^i_{jk} for its Christoffel symbols and RlkijR^l{}_{kij} for the coordinate curvature components of their connection. At a coordinate point xx, for coordinate vectors uu and vv, the connection term has components Γ(x)(u,v)i=∑j,kΓjki(x)ujvk\Gamma(x)(u,v)^i=\sum_{j,k}\Gamma^i_{jk}(x)u^jv^k.✦

A field VV along a curve γ:ℝ→M\gamma:\mathbb R\to M is an assignment of a vector V(s)∈Tγ(s)MV(s)\in T_{\gamma(s)}M to each time ss; it is smooth on a time set UU if its coordinates dcγ(s)V(s)dc_{\gamma(s)}V(s) are smooth on U∩γ−1(dom⁡c)U\cap\gamma^{-1}(\operatorname{dom}c) in every chart cc.✦ For x=c∘γx=c\circ\gamma and w(s)=dcγ(s)V(s)w(s)=dc_{\gamma(s)}V(s), the covariant derivative DsVD_sV of a field VV along γ\gamma is the field given by (13). The curvature term R(V,γ̇)γ̇R(V,\dot\gamma)\dot\gamma is obtained by applying the coordinate curvature operator to w,x′,x′w,x',x' and transporting the result by the inverse chart differential.✦ DsV=dcx(s)−1(w′(s)+Γ(x(s))(x′(s),w(s))).D_sV=dc^{-1}_{x(s)} \bigl(w'(s)+\Gamma(x(s))(x'(s),w(s))\bigr).(13)✦ An affinely parametrized gg-geodesic γ\gamma on UU is a curve in MM that is smooth on UU and has vanishing covariant acceleration there.✦

For fields V,WV,W along γ\gamma, the intrinsic index density ii and index form II are defined in (14); KK denotes a time interval, and the unrestricted index is the integral over ℝ\mathbb R. ig,γ(V,W;s)=g(DsV,DsW)−g(R(V,γ̇)γ̇,W),IK(V,W)=∫Kig,γ(V,W;s)ds,I(V,W)=∫ℝig,γ(V,W;s)ds.\begin{aligned} i_{g,\gamma}(V,W;s) &=g(D_sV,D_sW) -g\bigl(R(V,\dot\gamma)\dot\gamma,W\bigr),\\ I_K(V,W)&=\int_K i_{g,\gamma}(V,W;s)\,ds,& I(V,W)&=\int_{\mathbb R}i_{g,\gamma}(V,W;s)\,ds. \end{aligned}(14) Let 𝒟γ(U)\mathcal D_\gamma(U) be the space of fields along γ\gamma that are smooth on UU, have compact support, and have closed support contained in UU. A nonnegative index form II on UU is one whose quadratic values on 𝒟γ(U)\mathcal D_\gamma(U) are nonnegative.✦ For a unit-speed geodesic, the normal index IK⟂(V)I_K^\perp(V) is IK(V,V)I_K(V,V) minus the integral over KK of the squared derivative of g(V,γ̇)g(V,\dot\gamma).

The Claim

Theorem 1 (All neck passages). Let 0<δ≤1/20<\delta\leq1/2 and C≥18C\geq18. Assume the passage cometric bounds (10) and the smallness condition (12) with coefficients Ce=3CC_e=3C and Cm=Cm(C)C_m=C_m(C).

Let U⊂ℝU\subset\mathbb R be open, let γ\gamma be an affinely parametrized gδg_\delta-geodesic on UU, and let VV be a smooth field along γ\gamma on UU. Suppose a<ba<b, [a,b]⊂U[a,b]\subset U, and γ((a,b))⊂N\gamma((a,b))\subset N. At each endpoint e∈{a,b}e\in\{a,b\}, suppose either γ(e)∉N\gamma(e)\notin N or VV vanishes on a neighborhood of ee. Then I[a,b](V,V)≥0.I_{[a,b]}(V,V)\geq0.(15)✦

Let UU instead be an open interval, possibly empty or unbounded, and let γ\gamma be an affinely parametrized gδg_\delta-geodesic on UU. Then I(V,V)≥0for every V∈𝒟γ(U).I(V,V)\geq0\qquad\text{for every }V\in\mathcal D_\gamma(U).(16)✦

In the paper, the result reads: choose δ>0\delta>0 satisfying Lemma 3.1 and δ≤min⁡{1,1/(2Cm),(1−β)/(2Cm+Cea(0))}\delta\leq\min\{1,1/(2C_m),(1-\beta)/(2C_m+C_ea(0))\}, then choose the boundary periods as in Section 3; for every complete unit-speed geodesic γ\gamma of the resulting closed manifold, every connected component KK of its visit to the open neck, and every smooth compactly supported field VV along γ\gamma, both IK⟂(V)≥0I_K^\perp(V)\geq0 and IK(V,V)≥0I_K(V,V)\geq0.

What Is Checked

The finite assertion (15) and the whole-domain assertion (16) are stated over the glued smooth metric gδg_\delta, the two endpoint cometrics, and their common passage and smallness bounds. The first uses a finite segment with the specified endpoint conditions; the second uses an open interval of times and compactly supported fields with closed support in that domain.

The quotient projections for the circle, surface, and cylinder are smooth covering maps with deck orbits as fibers and smooth local inverses. Circle and cylinder deck translations are smooth isometries, as is the effective surface action on the hyperbolic half-plane.

Metric descent and metric gluing give smooth Riemannian inner products with bounded unit balls. The shear cutoff and the triangular frame are smooth, and the induced frame metric depends smoothly on position.✦

The radial map and its lift are smooth and commute with cylinder projection; for Λ≥4\Lambda\geq4, its differential multiplies the height component by the positive radial slope and leaves transverse components fixed.✦ The reversing cylinder map and its inverse are smooth.

The matching period is positive✦, and Λ−≥4\Lambda_-\geq4.✦ The left and right cylindrical metrics have the cusp formula (8) below height 1/41/4 with scales c−c_- and c+c_+. The core and cusp metric domains cover each complete piece✦ and their metrics agree on overlaps.✦ The piece charts cover MM✦, their inverse maps are smooth with invertible differentials, and the truncated piece metrics agree on their overlaps.✦

The product cusp chart is smooth and has an invertible differential for every cusp threshold greater than 11. The logarithmic cylinder coordinates and their inverses are smooth for every positive threshold, and the inclusion of the truncated piece into Σ×𝕊\Sigma\times\mathbb S is smooth with an invertible differential. Above any threshold greater than 11, two upper-half-plane points have the same surface quotient projection exactly when they have the same imaginary part and the same real coordinate modulo 66.✦