OctoInformBeta

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

Proposition 5.1 [Uniform returning-passage estimate]: Deep returns with arbitrary endpoint values in both frames, and intrinsic shallow and turning estimates

Uniform index estimates for returning 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

A returning neck passage enters at one boundary and leaves through that boundary again. The question is the sign of its normal index for a small positive neck parameter δ\delta.

Use coordinates (t,x,y)(t,x,y) on ℝ3\mathbb R^3 and the axial and transverse splitting U=(Ut,U⟂)∈ℝ×ℝ2U=(U_t,U_\perp)\in\mathbb R\times\mathbb R^2. For transverse vectors u,w,qu,w,q and a real 2×22\times2 matrix BB, the matrix bilinear pairing, its quadratic form, and the transverse squared norm have the conventions ℬB(u,w)=u𝖳Bw,ℬB(q,q)=q𝖳Bq,|q|2=q12+q22.\mathcal B_B(u,w)=u^{\mathsf T}Bw,\qquad \mathcal B_B(q,q)=q^{\mathsf T}Bq,\qquad |q|^2=q_1^2+q_2^2.(1)

The Neck and Its Endpoint Frames

Let θ\theta be the standard smooth transition from zero to one. The one-sided cutoff aa and its integral AA, together with the shear transition χ\chi and collar transition λ\lambda, are the functions a(t)={e−1/(2/3−t),t<2/3,0,t≥2/3,A(t)=∫0ta(u)du,χ(t)=θ(6(t−5/12)),λ(t)=θ(2t+3/2).\begin{aligned} a(t)&=\begin{cases}e^{-1/(2/3-t)},&t<2/3,\\0,&t\geq2/3,\end{cases} & A(t)&=\int_0^t a(u)\,du,\\ \chi(t)&=\theta\bigl(6(t-5/12)\bigr), &\lambda(t)&=\theta(2t+3/2). \end{aligned}(2) In terms of (2), the diagonal coefficients d1,d2d_1,d_2, shear σ\sigma, and cometric PP are defined by d1(δ,t)=1+δA(t),d2(δ,t)=1+δA(1−t),σ(δ,t)=δ2χ(t),P(δ,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) The entries in the matrix in (3) are evaluated at (δ,t)(\delta,t).

The endpoint frame matrix C2C_2 gives the right endpoint cometric. For the frame choice r∈{L,R}r\in\{\mathrm L,\mathrm R\} and inward endpoint height τ\tau, the coordinate cometric and its inverse metric matrix are C2(δ)=(011−δ2),PL,δ(τ)=P(δ,τ),PR,δ(τ)=C2(δ)P(δ,1−τ)C2(δ)𝖳,Qr,δ(τ)=Pr,δ(τ)−1.\begin{aligned} C_2(\delta)&=\begin{pmatrix}0&1\\1&-\delta^2\end{pmatrix},\\ P_{\mathrm L,\delta}(\tau)&=P(\delta,\tau),& P_{\mathrm R,\delta}(\tau) &=C_2(\delta)P(\delta,1-\tau)C_2(\delta)^{\mathsf T},\\ Q_{r,\delta}(\tau)&=P_{r,\delta}(\tau)^{-1}. \end{aligned}(4) Primes on the matrices in (4) denote differentiation with respect to inward height.

The collar denominator is d1(δ,t)d_1(\delta,t).✦ The terminal decay rate ρδ\rho_\delta, logarithmic decay kδk_\delta, and warping function ℓδ\ell_\delta are defined by ρδ(t)=δa(t)2d1(δ,t),kδ(t)=1−λ(t)+λ(t)ρδ(t),ℓδ(t)=exp⁡(−∫0tkδ(u)du).\rho_\delta(t)=\frac{\delta a(t)}{2d_1(\delta,t)},\qquad k_\delta(t)=1-\lambda(t)+\lambda(t)\rho_\delta(t),\qquad \ell_\delta(t)=\exp\left(-\int_0^t k_\delta(u)\,du\right).(5)

On ℝ3\mathbb R^3, a smooth family qzq_z of symmetric positive definite bilinear forms with bounded unit balls defines the Riemannian metric gz=qzg_z=q_z.✦ A smooth moving linear frame KK is a smooth family of linear automorphisms of the coordinate space. The metric gKg_K associated with a smooth moving linear frame KK on ℝ3\mathbb R^3 is the pullback of the Cartesian bilinear form ⟨u,w⟩=∑iuiwi\langle u,w\rangle=\sum_i u_iw_i by K(z)K(z).✦ The triangular frame 𝒯t(u)=(ut,α(t)(ux−c(t)uy),η(t)uy)\mathcal T_t(u)=(u_t,\alpha(t)(u_x-c(t)u_y),\eta(t)u_y), for smooth nonvanishing α,η\alpha,\eta and smooth cc, defines the cylinder metric gK(u,w)=⟨Ku,Kw⟩,g𝒯=dt2+α(t)2(dx−c(t)dy)2+η(t)2dy2.g_K(u,w)=\langle K u,K w\rangle,\qquad g_{\mathcal T}=dt^2+\alpha(t)^2(dx-c(t)\,dy)^2+\eta(t)^2dy^2.(6) For 0<δ≤1/20<\delta\leq1/2, the normalized neck metric gδg_\delta is the cylinder metric (6) with α(t)=ℓδ(t)\alpha(t)=\ell_\delta(t), η(t)=ℓδ(1−t)\eta(t)=\ell_\delta(1-t) from (5), and c(t)=σ(δ,t)c(t)=\sigma(\delta,t).✦

The right collar coordinate map fδf_\delta is the affine map whose linear part FδF_\delta is the isomorphism Fδ(τ,x,y)=(−τ,δ2x+y,x),Fδ−1(t,x,y)=(−t,y,x−δ2y),fδ(τ,x,y)=Fδ(τ,x,y)+(1,0,0).\begin{aligned} F_\delta(\tau,x,y)&=(-\tau,\delta^2x+y,x),\\ F_\delta^{-1}(t,x,y)&=(-t,y,x-\delta^2y),\\ f_\delta(\tau,x,y)&=F_\delta(\tau,x,y)+(1,0,0). \end{aligned}(7) For a Riemannian metric gg and a smooth map ff with invertible differential, the pullback metric f*gf^*g is the metric with inner product (f*g)z(u,w)=gf(z)(dfzu,dfzw)(f^*g)_z(u,w)=g_{f(z)}(df_z u,df_z w). The endpoint-framed neck metrics, using (7), are gL,δ=gδ,gR,δ=fδ*gδ.g_{\mathrm L,\delta}=g_\delta,\qquad g_{\mathrm R,\delta}=f_\delta^*g_\delta.(8)✦

The Index and Returning Data

For a smooth Riemannian metric gg on a three-manifold and a curve γ\gamma, the velocity is γ̇(s)=dγs(1)\dot\gamma(s)=d\gamma_s(1).✦ A field JJ along γ\gamma is smooth on a time set UU if its coordinates wc(s)=dcγ(s)J(s)w_c(s)=dc_{\gamma(s)}J(s) are smooth on U∩γ−1(dom⁡c)U\cap\gamma^{-1}(\operatorname{dom}c) for every atlas chart cc.✦

In a chart, let eie_i be the standard basis vectors, gij=g(dc−1ei,dc−1ej)g_{ij}=g(dc^{-1}e_i,dc^{-1}e_j) the metric matrix, and ∂if=df(ei)\partial_i f=df(e_i) the coordinate derivative. With (gij)=(gij)−1(g^{ij})=(g_{ij})^{-1}, the Christoffel symbols and curvature components are Γjki=12∑lgil(∂jglk+∂kglj−∂lgjk),Rlkij=∂iΓjkl−∂jΓikl+∑a(ΓialΓjka−ΓjalΓika).\begin{aligned} \Gamma^i_{jk} &=\frac12\sum_l g^{il} (\partial_jg_{lk}+\partial_kg_{lj}-\partial_lg_{jk}),\\ R^l{}_{kij} &=\partial_i\Gamma^l_{jk}-\partial_j\Gamma^l_{ik} +\sum_a(\Gamma^l_{ia}\Gamma^a_{jk} -\Gamma^l_{ja}\Gamma^a_{ik}). \end{aligned}(9) The connection operator Γ\Gamma and curvature operator RR are the operators with components Γ(x)(u,w)i=∑j,kΓjki(x)ujwk\Gamma(x)(u,w)^i=\sum_{j,k}\Gamma^i_{jk}(x)u^jw^k and (R(u,v)w)l=∑k,i,jRlkijwkuivj(R(u,v)w)^l=\sum_{k,i,j}R^l{}_{kij}w^ku^iv^j, respectively, with the coefficients of (9). The covariant derivative DsJD_sJ of a field JJ along a curve γ\gamma is the tangent vector at γ(s)\gamma(s) whose coordinates are wc′(s)+Γ(c(γ(s)))((c∘γ)′(s),wc(s))w_c'(s)+\Gamma(c(\gamma(s)))((c\circ\gamma)'(s),w_c(s)).✦ A geodesic γ\gamma on a time set UU in a Riemannian manifold is a smooth curve on UU with vanishing covariant acceleration in every chart.✦ For a field JJ along γ\gamma, the curvature term at time ss is the tangent vector R(J(s),γ̇(s))γ̇(s)R(J(s),\dot\gamma(s))\dot\gamma(s) at γ(s)\gamma(s).

For fields J,HJ,H along a curve γ\gamma in a Riemannian manifold with metric gg, the intrinsic index density igi_g is the scalar function and the intrinsic index IgI_g is its integral given by (10): ig(J,H;s)=g(DsJ,DsH)−g(R(J,γ̇)γ̇,H),Ig(J,H;[a,b])=∫abig(J,H;s)ds.\begin{aligned} i_g(J,H;s) &=g(D_sJ,D_sH)-g(R(J,\dot\gamma)\dot\gamma,H),\\ I_g(J,H;[a,b])&=\int_a^b i_g(J,H;s)\,ds. \end{aligned}(10) The metric and vectors in (10) are evaluated at γ(s)\gamma(s).

For transverse matrices P,Q,H,BP,Q,H,B and split vectors U,W,Z∈ℝ×ℝ2U,W,Z\in\mathbb R\times\mathbb R^2, let gQg_Q be the split metric and let Γ=ΓP,H\Gamma=\Gamma_{P,H} be the neck Christoffel operator, with axial derivative Γt=Γt;P,H,B\Gamma_t=\Gamma_{t;P,H,B}, given by gQ(U,W)=UtWt+ℬQ(U⟂,W⟂),Γ(U,W)t=−12ℬH(U⟂,W⟂),Γ(U,W)⟂=12(UtPHW⟂+WtPHU⟂),(Γt(U,W))t=−12ℬB(U⟂,W⟂),(Γt(U,W))⟂=12(Ut(PB−PHPH)W⟂+Wt(PB−PHPH)U⟂).\begin{aligned} g_Q(U,W)&=U_tW_t+\mathcal B_Q(U_\perp,W_\perp),\\ \Gamma(U,W)_t&=-\frac12\mathcal B_H(U_\perp,W_\perp),\\ \Gamma(U,W)_\perp&=\frac12(U_tPHW_\perp+W_tPHU_\perp),\\ \bigl(\Gamma_t(U,W)\bigr)_t &=-\frac12\mathcal B_B(U_\perp,W_\perp),\\ \bigl(\Gamma_t(U,W)\bigr)_\perp &=\frac12\bigl(U_t(PB-PHPH)W_\perp +W_t(PB-PHPH)U_\perp\bigr). \end{aligned}(11) The split coordinate curvature tensor formed from (11) is RP,H,B(U,W)Z=UtΓt(W,Z)−WtΓt(U,Z)+Γ(U,Γ(W,Z))−Γ(W,Γ(U,Z)).\begin{aligned} R_{P,H,B}(U,W)Z ={}&U_t\Gamma_t(W,Z)-W_t\Gamma_t(U,Z)\\ &+\Gamma(U,\Gamma(W,Z))-\Gamma(W,\Gamma(U,Z)). \end{aligned}(12)✦ The normal index density 𝒩P,Q,H,B\mathcal N_{P,Q,H,B} is the scalar function of split vectors T,V,DT,V,D given by (13), with Ξ=D+Γ(T,V)\Xi=D+\Gamma(T,V) and curvature (12): 𝒩P,Q,H,B(T,V,D)=gQ(Ξ,Ξ)−gQ(V,RP,H,B(V,T)T)−gQ(Ξ,T)2.\mathcal N_{P,Q,H,B}(T,V,D) =g_Q(\Xi,\Xi)-g_Q(V,R_{P,H,B}(V,T)T)-g_Q(\Xi,T)^2.(13)✦

Fix an endpoint frame rr, momentum p∈ℝ2p\in\mathbb R^2, height t(s)t(s) and height velocity v(s)v(s). The momentum potential 𝒱p\mathcal V_p, reduced force FpF_p, and velocity at fixed momentum TT are 𝒱p(τ)=p𝖳Pr,δ(τ)p,Fp(τ)=p𝖳Pr,δ′(τ)p,T(s)=(v(s),Pr,δ(t(s))p).\begin{aligned} \mathcal V_p(\tau)&=p^{\mathsf T}P_{r,\delta}(\tau)p, &F_p(\tau)&=p^{\mathsf T}P'_{r,\delta}(\tau)p,\\ T(s)&=\bigl(v(s),P_{r,\delta}(t(s))p\bigr). \end{aligned}(14) For paths V,DV,D in the split tangent space, the returning density nr(s;V,D)n_r(s;V,D) is (13) with the matrices evaluated along t(s)t(s) and with TT from (14).✦ The normal index I⟂(V)I^\perp(V) of a split field VV on a passage of duration LL is the integral in (15): nr(s;V,D)=𝒩P(t(s)),Q(t(s)),Q′(t(s)),Q″(t(s))(T(s),V(s),D(s)),I⟂(V)=∫0Lnr(s;V,V′)ds,P=Pr,δ,Q=Qr,δ.\begin{aligned} n_r(s;V,D) &=\mathcal N_{P(t(s)),Q(t(s)),Q'(t(s)),Q''(t(s))} (T(s),V(s),D(s)),\\ I^\perp(V)&=\int_0^L n_r(s;V,V')\,ds, \qquad P=P_{r,\delta},\quad Q=Q_{r,\delta}. \end{aligned}(15)

The Turning Parameters

For a real coefficient CC, the turning constants a0,βa_0,\beta and the mixed coefficient bound m(C)m(C) are a0=e−3/2,β=1+e−9/262,m(C)=a(0)+2Ca(0)+22Ca(0)3.a_0=e^{-3/2},\qquad \beta=\frac{1+e^{-9/26}}2,\qquad m(C)=a(0)+\sqrt{2Ca(0)}+2\sqrt{\frac{2Ca(0)}3}.(16) For real coefficients Ce,CmC_e,C_m, the turning smallness range 𝒟(Ce,Cm)\mathcal D(C_e,C_m) is the set in (17), using the constants of (16): 𝒟(Ce,Cm)={δ:0<δ≤1,δ≤12Cm,δ≤1−β2Cm+Cea0},\mathcal D(C_e,C_m)= \left\{\delta:\ 0<\delta\leq1,\quad \delta\leq\frac1{2C_m},\quad \delta\leq\frac{1-\beta}{2C_m+C_ea_0}\right\},(17)✦

For an endpoint frame rr, coefficient CC and neck parameter δ\delta, the full cometric bounds ℋr(C,δ)\mathcal H_r(C,\delta) are the condition on Pr,δP_{r,\delta} that the following inequalities hold for every τ∈[0,1]\tau\in[0,1] and q∈ℝ2q\in\mathbb R^2: |q|29≤q𝖳Pr,δ(τ)q≤18|q|2,|Pr,δ′(τ)q|2≤C2δ2|q|2,0≤−q𝖳Pr,δ″(τ)q,|Pr,δ′(τ)q|2≤Cδ(−q𝖳Pr,δ″(τ)q).\begin{aligned} \frac{|q|^2}{9} &\leq q^{\mathsf T}P_{r,\delta}(\tau)q\leq18|q|^2,\\ |P'_{r,\delta}(\tau)q|^2&\leq C^2\delta^2|q|^2,\\ 0&\leq-q^{\mathsf T}P''_{r,\delta}(\tau)q,\\ |P'_{r,\delta}(\tau)q|^2 &\leq C\delta\bigl(-q^{\mathsf T}P''_{r,\delta}(\tau)q\bigr). \end{aligned}(18)

The Claim

Theorem 1. There are constants 0<δ0≤1/20<\delta_0\leq1/2 and C≥18C\geq18 such that every 0<δ≤δ00<\delta\leq\delta_0, in each endpoint frame rr, satisfies δ∈𝒟(3C,m(C)),ℋr(C,δ).\delta\in\mathcal D(3C,m(C)),\qquad \mathcal H_r(C,\delta).(19)✦

For the deep returning estimate, fix a frame rr, momentum p∈ℝ2p\in\mathbb R^2, C≥18C\geq18 and δ∈𝒟(3C,m(C))\delta\in\mathcal D(3C,m(C)). Assume the comparison and concavity bounds of (18). Let e=(1,0)e=(1,0) be the first tangential basis vector, and assume also that |Pr,δ′(τ)e|2≤C2δ2|P'_{r,\delta}(\tau)e|^2\leq C^2\delta^2 and |Pr,δ′(τ)p|2≤Cδ(−p𝖳Pr,δ″(τ)p)|P'_{r,\delta}(\tau)p|^2\leq C\delta(-p^{\mathsf T}P''_{r,\delta}(\tau)p) for every τ∈[0,1]\tau\in[0,1]. Let L>0L>0 and let t,v:ℝ→ℝt,v:\mathbb R\to\mathbb R be smooth at every time of [0,L][0,L], with t′=vt'=v, v′=−Fp(t)/2v'=-F_p(t)/2 and 𝒱p(t)+v2=1\mathcal V_p(t)+v^2=1 there. Suppose t(0)=t(L)=0t(0)=t(L)=0, 0<t(s)<10<t(s)<1 for 0<s<L0<s<L, and t(L/2)>1/4t(L/2)>1/4. For every split field VV smooth at every time of [0,L][0,L], with arbitrary endpoint values, the normal index (15) satisfies I⟂(V)≥0.I^\perp(V)\geq0.(20)✦

In the shallow estimate, fix a frame rr, 0<δ≤1/20<\delta\leq1/2 and a≤ba\leq b. Let γ:ℝ→ℝ3\gamma:\mathbb R\to\mathbb R^3 be smooth at every time in [a,b][a,b] and have height in [0,1/4][0,1/4] throughout that interval. For every field J:ℝ→ℝ3J:\mathbb R\to\mathbb R^3, the intrinsic index for (8) satisfies Igr,δ(J,J;[a,b])≥0.I_{g_{r,\delta}}(J,J;[a,b])\geq0.(21)✦

Fix 0<δ≤1/20<\delta\leq1/2, C≥18C\geq18 and a frame rr for the turning estimate, with δ∈𝒟(3C,m(C))\delta\in\mathcal D(3C,m(C)) and ℋr(C,δ)\mathcal H_r(C,\delta). Let L>0L>0 and let U⊂ℝU\subset\mathbb R be open with [0,L]⊂U[0,L]\subset U. Let γ\gamma be a geodesic for gr,δg_{r,\delta} on UU, and let JJ be a smooth field along it on UU. Suppose γ\gamma has unit speed on [0,L][0,L] and constant transverse momentum Qr,δ(t(s))γ̇⟂(s)=pQ_{r,\delta}(t(s))\dot\gamma_\perp(s)=p there, where t(s)=γ0(s)t(s)=\gamma_0(s). Assume t(0)=0t(0)=0, t(L)∈[0,1]t(L)\in[0,1], 0<t(s)<10<t(s)<1 for 0<s<L0<s<L, and t′(c)=0t'(c)=0 at some c∈(0,L)c\in(0,L). Suppose either t(L)∈{0,1}t(L)\in\{0,1\} or JJ vanishes on a neighborhood of LL. Then Igr,δ(J,J;[0,L])≥0.I_{g_{r,\delta}}(J,J;[0,L])\geq0.(22)✦

In the paper, the result reads: for every sufficiently small positive δ\delta, every returning geodesic passage of the constructed neck has I⟂(V)≥0I^\perp(V)\geq0 for every smooth field VV along the closed passage with unrestricted endpoint values, with one choice of δ\delta valid at both boundaries for all unit-speed returning geodesics.

What Is Checked

The deep-return statement (20) is over the split normal density and smooth solutions of the fixed-momentum height equations; the uniform parameter statement (19) is over the two constructed coordinate cometrics. The intrinsic statements (21) and (22) are over the smooth endpoint-framed metrics, their covariant derivatives and curvature, with the stated interval and endpoint conditions.

For smooth radius functions that never vanish and smooth shear, the triangular frame varies smoothly with the point.✦ Pulling back the Cartesian bilinear form through a smooth moving frame gives a smooth bilinear field✦. The pullback bilinear field of a Riemannian metric by a smooth map is a smooth section.✦