OctoInformBeta

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

§5 (setting): the paper's dual form and the explicit replacement: The completed-square energy density and endpoint identity for transverse variations

Endpoint coefficients as an explicit quadratic.

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 variation along a returning geodesic has a transverse component z(s)z(s) and a height component w(s)w(s), where ss is geodesic time. The endpoint functional pairs the transverse displacement with a fixed momentum direction.

Transverse vectors belong to ℝ2\mathbb R^2✦, and transverse matrices belong to ℝ2×2\mathbb R^{2\times2}.✦ The matrix bilinear pairing associated with A∈ℝ2×2A\in\mathbb R^{2\times2} is the real bilinear form B(A;⋅,⋅)B(A;\cdot,\cdot) defined by (1): B(A;u1,u2)=u1𝖳Au2,u1,u2∈ℝ2.B(A;u_1,u_2)=u_1^{\mathsf T}Au_2, \qquad u_1,u_2\in\mathbb R^2.(1)✦

The transverse cometric along the geodesic is P(s)P(s), with metric matrix Q(s)=P(s)−1Q(s)=P(s)^{-1}. Write p∈ℝ2p\in\mathbb R^2 for the conserved momentum, e∈ℝ2e\in\mathbb R^2 for a fixed transverse direction, and v(s)v(s) for the height velocity. A vector path k(s)k(s) specifies the coupling to the height variation.

The Explicit Coefficients

For a return duration L≥0L\geq0 and a constant K∈ℝK\in\mathbb R, the explicit scalar coefficients spp\mathrm{spp}, spe\mathrm{spe}, and see\mathrm{see} are the quantities in (2), whenever their integrands are integrable: spp=∫0LB(P(s);p,p)ds+4∫0Lv(s)2ds,spe=∫0LB(P(s);e,p)ds,see=∫0LB(P(s);e,e)ds+KL.\begin{aligned} \mathrm{spp}&=\int_0^L B(P(s);p,p)\,ds +4\int_0^L v(s)^2\,ds,\\ \mathrm{spe}&=\int_0^L B(P(s);e,p)\,ds,\\ \mathrm{see}&=\int_0^L B(P(s);e,e)\,ds+KL. \end{aligned}(2)

The Claim

Definition 1 (Completed-square energy density). Let QQ be a real 2×22\times2 matrix path, kk and ζ\zeta vector paths in ℝ2\mathbb R^2, and h,w,ηh,w,\eta real functions. The completed-square energy density ε\varepsilon is the real function defined by (3): ε(s)=B(Q(s);ζ(s)−w(s)k(s),ζ(s)−w(s)k(s))+η(s)2+h(s)w(s)2.\varepsilon(s) =B\bigl(Q(s);\zeta(s)-w(s)k(s),\zeta(s)-w(s)k(s)\bigr) +\eta(s)^2+h(s)w(s)^2.(3)✦

Theorem 1 (Endpoint identity and coefficient expansions). Let a,b∈ℝa,b\in\mathbb R, let p,e∈ℝ2p,e\in\mathbb R^2, and let α∈ℝ\alpha\in\mathbb R. Suppose z:ℝ→ℝ2z:\mathbb R\to\mathbb R^2 and v,w:ℝ→ℝv,w:\mathbb R\to\mathbb R are differentiable at every point of [min⁡{a,b},max⁡{a,b}][\min\{a,b\},\max\{a,b\}], with w(a)=w(b)=0w(a)=w(b)=0 and p⋅k(s)=−2v̇(s)p\cdot k(s)=-2\dot v(s) throughout that interval. If the integrand in (4) is integrable there, the endpoint functional satisfies (4): $$\begin{aligned} &\int_a^b\Bigl[ (p-\alpha e)\cdot\bigl(\dot z(s)-w(s)k(s)\bigr) +2v(s)\dot w(s)-\alpha\bigl(e\cdot k(s)\bigr)w(s) \Bigr],ds\\ &\hspace{25mm}=(p-\alpha e)\cdot\bigl(z(b)-z(a)\bigr). \end{aligned}$$(4)✦

Let PP be a symmetric real matrix path and K∈ℝK\in\mathbb R. Suppose B(P(s);p,p)B(P(s);p,p), B(P(s);p,e)B(P(s);p,e), B(P(s);e,e)B(P(s);e,e), and v(s)2v(s)^2 are integrable between aa and bb. The direct coefficient expansion is (5): ∫ab[B(P(s);p−αe,p−αe)+4v(s)2+α2K],ds=∫ab[B(P(s);p,p)+4v(s)2],ds−2α∫abB(P(s);p,e)ds+α2∫ab[B(P(s);e,e)+K],ds.\begin{aligned} &\int_a^b\Bigl[B(P(s);p-\alpha e,p-\alpha e) +4v(s)^2+\alpha^2K\Bigr],ds\\ &\quad=\int_a^b\Bigl[B(P(s);p,p)+4v(s)^2\Bigr],ds -2\alpha\int_a^b B(P(s);p,e)\,ds\\ &\qquad\quad+\alpha^2\int_a^b\Bigl[B(P(s);e,e)+K\Bigr],ds. \end{aligned}(5)✦

For L≥0L\geq0, suppose PP and vv are continuous on [0,L][0,L] and P(s)P(s) is symmetric there. With the coefficients of (2), the returning coefficient expansion is (6): ∫0L[B(P(s);p−αe,p−αe)+4v(s)2+α2K],ds=spp−2αspe+α2see.\int_0^L\Bigl[B(P(s);p-\alpha e,p-\alpha e) +4v(s)^2+\alpha^2K\Bigr],ds =\mathrm{spp}-2\alpha\,\mathrm{spe}+\alpha^2\mathrm{see}.(6)✦

In the paper, the result reads: for a gauged return, with scalar κ\kappa, initial height velocity v0=v(0)v_0=v(0), and a function FF, with primes on PP and FF denoting height derivatives, put ξ=ż−P′pw\xi=\dot z-P'pw, X=z(L)−z(0)X=z(L)-z(0), and μ=κv0/4\mu=\kappa v_0/4; the space D=H01(0,L)D=H_0^1(0,L) has inner product ⟨w1,w2⟩D=int0L(ẇ1ẇ2−F″w1w2/2)ds\langle w_1,w_2\rangle_D=int_0^L (\dot w_1\dot w_2-F''w_1w_2/2)\,ds, and, for q,r∈ℝ2q,r\in\mathbb R^2, with ⟨⋅,⋅⟩D*\langle\cdot,\cdot\rangle_{D^*} denoting the dual inner product, the functionals Φq(w)=∫0Lq𝖳P′pwds\Phi_q(w)=\int_0^L q^{\mathsf T}P'pw\,ds give the positive definite form S(q,r)=∫0Lq𝖳Prds+⟨Φq,Φr⟩D*S(q,r)=\int_0^L q^{\mathsf T}Pr\,ds+ \langle\Phi_q,\Phi_r\rangle_{D^*}, whose diagonal gives the squared norm of the endpoint functional on the Hilbert direct sum of the ξ\xi and ww spaces.

What Is Checked

The energy density is stated over arbitrary real matrix, vector, and scalar paths; the endpoint identity is stated over differentiable variations with vanishing height endpoints and the relation p⋅k=−2v̇p\cdot k=-2\dot v, while the coefficient identities are stated over symmetric matrix paths with the indicated integrability or continuity conditions.