OctoInformBeta

OpenAI preprint, September 2026 · Section 5. Preparation, whitening, and the smooth limit

Section 5: retaining polynomials and taking the smooth limit: Prescribed approximation errors and summable jet changes for real centered vectors on 𝕋3\mathbb T^3

Retaining orbit polynomials through smooth limits.

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 approximation problem is to retain finite orbit polynomials while conjugated shears and their generating vectors change. The dynamical maps live on 𝕋3\mathbb T^3, where 𝕋=ℝ/ℤ\mathbb T=\mathbb R/\mathbb Z, and, writing μ\mu for normalized Haar measure on 𝕋3\mathbb T^3, the vectors live in the complex Hilbert space H=L2(𝕋3,μ)H=L^2(\mathbb T^3,\mu). Here λ\lambda is normalized Haar measure on 𝕋\mathbb T, the slow-coordinate measure is λ2\lambda^2, and μ=λ⊗λ2\mu=\lambda\otimes\lambda^2. The coordinate and product descriptions of the three-torus are identified by (x0,x1,x2)↦(x0,(x1,x2))(x_0,x_1,x_2)\mapsto(x_0,(x_1,x_2)).✦ The covering projection π:ℝ3→𝕋3\pi:\mathbb R^3\to\mathbb T^3 reduces each coordinate modulo one.✦

The volume integral ℐ:H→ℂ\mathcal I:H\to\mathbb C is the continuous complex linear functional in (1), and its kernel H0=L02(𝕋3,μ)H_0=L^2_0(\mathbb T^3,\mu) is the full complex mean-zero subspace: ℐ(g)=⟨1,g⟩H=∫𝕋3gdμ,H0=ker⁡ℐ.\mathcal I(g)=\langle 1,g\rangle_H=\int_{\mathbb T^3}g\,d\mu, \qquad H_0=\ker\mathcal I.(1) A real centered physical test hh is a continuous function h:𝕋3→ℂh:\mathbb T^3\to\mathbb C with h¯=h\overline{h}=h, ℐ(h)=0\mathcal I(h)=0, and |h(x)|≤1|h(x)|\leq1 everywhere.✦ Write [h][h] for its almost-everywhere class in HH.✦

Admissible Stages

A measurable packet family 𝒫\mathcal P consists of a finite index set II, distinct nonzero integers mim_i, and measurable square-integrable amplitudes Ai:𝕋2→ℂA_i:\mathbb T^2\to\mathbb C for λ2\lambda^2.✦ With e(t)=exp⁡(2πit)e(t)=\exp(2\pi i t) and coordinates (θ,z,a)∈𝕋3(\theta,z,a)\in\mathbb T^3, the angular packets pip_i and finite packet sum PP of 𝒫\mathcal P are defined by (2): pi(θ,z,a)=Ai(z,a)e(miθ),P(θ,z,a)=∑i∈Ipi(θ,z,a).p_i(\theta,z,a)=A_i(z,a)e(m_i\theta), \qquad P(\theta,z,a)=\sum_{i\in I}p_i(\theta,z,a).(2) For a circle phase map φ\varphi, the packet spectral measures and their family spectrum are the weighted pushforwards in (3): νi,φ=((z,a)↦miφ(z))*(|Ai(z,a)|2dλ(z)dλ(a)),ν𝒫,φ=∑i∈Iνi,φ.\nu_{i,\varphi} =\bigl((z,a)\mapsto m_i\varphi(z)\bigr)_* \bigl(|A_i(z,a)|^2\,d\lambda(z)\,d\lambda(a)\bigr), \qquad \nu_{\mathcal P,\varphi}=\sum_{i\in I}\nu_{i,\varphi}.(3) A measurable packet family 𝒫\mathcal P with index set II is reflection-symmetric if there is an involution τ:I→I\tau:I\to I satisfying mτ(i)=−mim_{\tau(i)}=-m_i and Aτ(i)=Ai¯A_{\tau(i)}=\overline{A_i} pointwise. A measurable packet family 𝒫\mathcal P is supported in a real interval [l,u][l,u] if each of its amplitudes has support in the product of [l,u]mod⁡ℤ[l,u]\bmod\mathbb Z with the second circle.✦

A physical coordinate frame FF is a bijection χF\chi_F of 𝕋3\mathbb T^3 with a smooth forward lift LF:ℝ3→ℝ3L_F:\mathbb R^3\to\mathbb R^3 and smooth backward lifts RF,k:ℝ3→ℝ3R_{F,k}:\mathbb R^3\to\mathbb R^3, k∈ℤk\in\mathbb Z, satisfying π∘LF=χF∘π\pi\circ L_F=\chi_F\circ\pi and π∘RF,k=χF−1∘π\pi\circ R_{F,k}=\chi_F^{-1}\circ\pi.✦ For a real speed VV and real time rr, the covering shear CVrC_V^r and the physical iterate lift ΦF,Vk\Phi_{F,V}^k are given by (4): CVr(x0,x1,x2)=(x0+rV(x1),x1,x2),ΦF,Vk=RF,k∘CVk∘LF(k∈ℤ).C_V^r(x_0,x_1,x_2)=(x_0+rV(x_1),x_1,x_2), \qquad \Phi_{F,V}^k=R_{F,k}\circ C_V^k\circ L_F \quad(k\in\mathbb Z).(4)

An admissible physical stage s=(F,𝒫,V,δ,l,u)s=(F,\mathcal P,V,\delta,l,u) is a tuple consisting of a physical coordinate frame FF with a continuous μ\mu-preserving chart χF\chi_F, a reflection-symmetric packet family 𝒫\mathcal P supported in [l,u][l,u], a smooth one-periodic real profile VV, and real parameters δ,l,u\delta,l,u, subject to (5). Here φs(z)=V(z)mod⁡ℤ\varphi_s(z)=V(z)\bmod\mathbb Z denotes the descended circle speed: 0<δ<12,δ<l<u<1−δ,V′(x)>0(δ<x<1−δ),ν𝒫,φs=λ.\begin{gathered} 0<\delta<\tfrac12,\qquad \delta<l<u<1-\delta, \qquad V'(x)>0\quad(\delta<x<1-\delta),\\ \nu_{\mathcal P,\varphi_s}=\lambda. \end{gathered}(5) The torus shear with speed φ\varphi is the bijection Θφ(θ,z,a)=(θ+φ(z),z,a)\Theta_\varphi(\theta,z,a)=(\theta+\varphi(z),z,a), with inverse Θ−φ\Theta_{-\varphi}.✦ The stage iterates TskT_s^k and physical packet vector fsf_s are defined by (6): Tsk=χF−1∘Θkφs∘χF(k∈ℤ),fs=[P∘χF]∈H.T_s^k=\chi_F^{-1}\circ\Theta_{k\varphi_s}\circ\chi_F \quad(k\in\mathbb Z), \qquad f_s=[P\circ\chi_F]\in H.(6)

Jet Restrictions and Orbit Polynomials

A finite physical jet restriction QQ is a finite list of integer iterates kik_i, derivative orders qiq_i, and compact sets Ki⊆ℝ3K_i\subseteq\mathbb R^3.✦ Two frame-speed pairs (F,V)(F,V) and (G,W)(G,W) are QQ-close within ε\varepsilon, written 𝒞Q(ε;(F,V),(G,W))\mathcal C_Q(\varepsilon;(F,V),(G,W)), when (7) holds: ∥DqiΦG,Wki(x)−DqiΦF,Vki(x)∥≤εfor every i and x∈Ki.\left\|D^{q_i}\Phi_{G,W}^{k_i}(x) -D^{q_i}\Phi_{F,V}^{k_i}(x)\right\|\leq\varepsilon \qquad\text{for every }i\text{ and }x\in K_i.(7)✦ Write 𝒞Q(ε;s,t)\mathcal C_Q(\varepsilon;s,t) for the same condition on two stages. The compact jet exhaustion QNQ_N uses the finite iterate and order indices JNJ_N in (8), on the indicated common domain: JN={−N,…,N}×{0,…,N},Kk,q=B¯(0,N+1)⊆ℝ3.J_N=\{-N,\ldots,N\}\times\{0,\ldots,N\}, \qquad K_{k,q}=\overline B(0,N+1)\subseteq\mathbb R^3.(8)

A smooth volume-preserving integer action D=(Dk)k∈ℤD=(D_k)_{k\in\mathbb Z} is a family of continuous μ\mu-preserving maps of 𝕋3\mathbb T^3 with D0=idD_0=\mathrm{id} and Dk+j=Dk∘DjD_{k+j}=D_k\circ D_j, admitting smooth lifts Ψk:ℝ3→ℝ3\Psi_k:\mathbb R^3\to\mathbb R^3 such that π∘Ψk=Dk∘π\pi\circ\Psi_k=D_k\circ\pi.✦ For T=D1T=D_1 and a vector f∈Hf\in H, the bilateral Koopman orbit of ff and its Laurent polynomials are defined by (9); the coefficient sequence c:ℤ→ℂc:\mathbb Z\to\mathbb C has finite support: UTkf=f∘Dk(k∈ℤ),Pc(UT)f=∑k∈supp⁡cckUTkf.U_T^k f=f\circ D_k\quad(k\in\mathbb Z), \qquad P_c(U_T)f=\sum_{k\in\operatorname{supp}c}c_k U_T^k f.(9) For a stage, the same notation uses TskT_s^k and fsf_s, including negative iterates.

The Claim

Theorem 1. Let (hi)i≥0(h_i)_{i\geq0} be real centered physical tests, and let (ei)(e_i) and (bi)(b_i) be strictly positive real sequences. There are admissible stages (sn)(s_n) and finitely supported coefficient sequences (ci)(c_i) with the bounds (10): ∥fsn+1−fsn∥H<bn,𝒞Qn(bn;sn,sn+1)(n≥0),∥[hi]−Pci(UTsn)fsn∥H<ei(0≤i<n).\begin{gathered} \|f_{s_{n+1}}-f_{s_n}\|_H<b_n, \qquad \mathcal C_{Q_n}(b_n;s_n,s_{n+1})\quad(n\geq0),\\ \|[h_i]-P_{c_i}(U_{T_{s_n}})f_{s_n}\|_H<e_i \qquad(0\leq i<n). \end{gathered}(10)✦

Let (sn)(s_n) be admissible stages and let (bn)(b_n) be a summable real sequence. Assume 𝒞Qn(bn;sn,sn+1)\mathcal C_{Q_n}(b_n;s_n,s_{n+1}) for every nn. There is a smooth volume-preserving integer action DD with the uniform convergence (11), where dd is the torus metric: limn→∞supx∈𝕋3d(Tsnkx,Dkx)=0(k∈ℤ).\lim_{n\to\infty}\sup_{x\in\mathbb T^3} d\bigl(T_{s_n}^k x,D_k x\bigr)=0 \qquad(k\in\mathbb Z).(11)✦

For any real centered physical tests (hi)(h_i) and strictly positive tolerances (ei)(e_i), there are a smooth volume-preserving integer action DD, a vector f∈Hf\in H, and finitely supported sequences (ci)(c_i) satisfying (12), with T=D1T=D_1 and δkj\delta_{kj} the Kronecker delta: f∈H0,Im⁡f=0μ-almost everywhere,⟨UTkf,UTjf⟩H=δkj(k,j∈ℤ),∥[hi]−Pci(UT)f∥H≤ei(i≥0).\begin{gathered} f\in H_0,\qquad \operatorname{Im}f=0\quad\mu\text{-almost everywhere},\\ \langle U_T^k f,U_T^j f\rangle_H=\delta_{kj}\quad(k,j\in\mathbb Z), \qquad \|[h_i]-P_{c_i}(U_T)f\|_H\leq e_i\quad(i\geq0). \end{gathered}(12)✦

There is a sequence (hi)(h_i) of real centered physical tests with the totality property (13) for every closed complex linear subspace KK of HH: ([hi]∈K for every i≥0)⇒H0⊆K.\bigl([h_i]\in K\text{ for every }i\geq0\bigr) \quad\Longrightarrow\quad H_0\subseteq K.(13)✦

In the paper, the result reads: repeated dense bounded real smooth centered tests are chosen together with successively small changes of the maps, their inverses and the vectors, and finite orbit polynomials whose approximations survive every later stage; the maps and inverses converge smoothly to mutually inverse volume-preserving diffeomorphisms and the vectors converge in HH, flat moments give an orthonormal bilateral orbit, and the retained dense tests make its closed span equal to H0H_0.

What Is Checked

The assertions are stated over admissible physical stages, finite jet restrictions on the universal cover, and real centered continuous tests in the complex space HH. Every stage iterate TskT_s^k is continuous✦ and preserves μ\mu✦; the physical packet function P∘χFP\circ\chi_F and every physical test are square-integrable.