OctoInformBeta

OpenAI preprint, September 2026 Β· Section 2. Smooth passages and spectral densities

Density comparison (2.4): Fixed angular modes and smooth passage families with a common twist

Packet proximity controls spectral density distance.

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

Finite angular Fourier sums on the three-torus have spectral measures determined by slow amplitudes and a circle-valued twist. The comparison concerns changes of these densities when the physical functions vary in L2L^2.

Write 𝕋=ℝ/β„€\mathbb T=\mathbb R/\mathbb Z and e(t)=exp⁡(2Ο€it)e(t)=\exp(2\pi i t). Let ΞΌ\mu be normalized Haar measure on 𝕋\mathbb T, both for angular coordinates and for the spectral circle.✦ The slow-coordinate Haar measure is ΞΌ2=ΞΌβŠ—ΞΌ\mu_2=\mu\otimes\mu on 𝕋2\mathbb T^2✦, and the product Haar measure is ΞΌ3=ΞΌβŠ—ΞΌ2\mu_3=\mu\otimes\mu_2 on 𝕋3\mathbb T^3.✦

For an integer mm and a complex amplitude AA on 𝕋2\mathbb T^2, the angular packet Pm,AP_{m,A} is the function on 𝕋3\mathbb T^3 given below, where ΞΈ\theta is the angular coordinate and y,ay,a are the slow coordinates. A finite index set ΞΉ\iota, an integer mode list m=(mi)i∈ιm=(m_i)_{i\in\iota} and amplitudes A=(Ai)i∈ιA=(A_i)_{i\in\iota} determine the finite packet sum Fm,AF_{m,A} on 𝕋3\mathbb T^3 in (1). Pm,A(ΞΈ,y,a)=A(y,a)e(mΞΈ),Fm,A(ΞΈ,y,a)=βˆ‘i∈ιAi(y,a)e(miΞΈ).P_{m,A}(\theta,y,a)=A(y,a)e(m\theta),\qquad F_{m,A}(\theta,y,a)=\sum_{i\in\iota}A_i(y,a)e(m_i\theta).(1) The squared physical error β„°(f,g)\mathcal E(f,g) and physical distance d(f,g)d(f,g) of complex functions f,gf,g on 𝕋3\mathbb T^3 are β„°(f,g)=βˆ«π•‹3|fβˆ’g|2dΞΌ3,d(f,g)=β„°(f,g).\mathcal E(f,g)=\int_{\mathbb T^3}|f-g|^2\,d\mu_3,\qquad d(f,g)=\sqrt{\mathcal E(f,g)}.(2)

Let Ο•:𝕋→𝕋\phi:\mathbb T\to\mathbb T be a measurable twist. The coefficient space for (1) is ι×𝕋2\iota\times\mathbb T^2. On the coefficient space ι×𝕋2\iota\times\mathbb T^2, the counting product Haar measure Ξ½\nu, the nonnegative real function EAE_A called the coefficient energy density, and the circle-valued function ΞΎΟ•,m\xi_{\phi,m} called the coefficient frequency are given by (3). Ξ½=#ΞΉβŠ—ΞΌ2,EA(i,y,a)=|Ai(y,a)|2,ΞΎΟ•,m(i,y,a)=miΟ•(y).\nu=\#_{\iota}\otimes\mu_2,\qquad E_A(i,y,a)=|A_i(y,a)|^2,\qquad \xi_{\phi,m}(i,y,a)=m_i\phi(y).(3) Here #ΞΉ\#_{\iota} denotes counting measure; the energy weight of any complex function is its squared absolute value. The finite packet spectral measure σϕ,m,A\sigma_{\phi,m,A} on 𝕋\mathbb T is the pushforward of the weighted coefficient measure EAdΞ½E_A\,d\nu under the coefficient frequency ΞΎΟ•,m\xi_{\phi,m}, as in (4). σϕ,m,A=(ΞΎΟ•,m)*(EAdΞ½).\sigma_{\phi,m,A}=(\xi_{\phi,m})_*(E_A\,d\nu).(4) A spectral density pp of (4) is a nonnegative measurable integrable function on 𝕋\mathbb T satisfying dσϕ,m,A=pdΞΌ.d\sigma_{\phi,m,A}=p\,d\mu.(5)✦

Smooth Passage Densities

A smooth passage family 𝒫\mathcal P consists of a finite list of nonzero integer modes mim_i, amplitudes Ai:𝕋2β†’β„‚A_i:\mathbb T^2\to\mathbb C, lifts A0,i:𝕋×ℝ→ℂA_{0,i}:\mathbb T\times\mathbb R\to\mathbb C, real parameters o,l,u,Ro,l,u,R, and a smooth real speed VV of period one. The amplitudes are continuous and smooth in both real covering coordinates. The lifts are jointly continuous, are twice continuously differentiable in the real coordinate with jointly continuous first derivative, and are smooth in both real covering coordinates. Their passage and matching conditions are o≀l≀u≀o+1,R>0,supp⁡A0,i(a,β‹…)βŠ†(l,u],Ai(xβ€Ύ,a)=A0,i(a,x)(o<x≀o+1),Vβ€²(x)>0(lβˆ’2R≀x≀u+2R),\begin{gathered} o\le l\le u\le o+1,\qquad R>0,\qquad \operatorname{supp}A_{0,i}(a,\cdot)\subseteq(l,u],\\ A_i(\bar x,a)=A_{0,i}(a,x)\quad(o<x\le o+1),\qquad V'(x)>0\quad(l-2R\le x\le u+2R), \end{gathered}(6) where xβ€Ύ\bar x is reduction modulo one.

For a set KβŠ†β„K\subseteq\mathbb R and real functions J,g,D,wJ,g,D,w on ℝ\mathbb R, the inverse branch density bK,J,g,D,w:ℝ→ℝb_{K,J,g,D,w}:\mathbb R\to\mathbb R is the real function given by (7). bK,J,g,D,w(s)={w(g(s))/|D(g(s))|,s∈J(K),0,sβˆ‰J(K).b_{K,J,g,D,w}(s)= \begin{cases} w(g(s))/|D(g(s))|,&s\in J(K),\\ 0,&s\notin J(K). \end{cases}(7)✦ Let gJ,K(s)g_{J,K}(s) select a preimage in KK under JJ when s∈J(K)s\in J(K). For a complex profile BB on ℝ\mathbb R, the stationary frequency density SK,J,BS_{K,J,B} is the real function obtained from the branch density (7) with derivative denominator and energy weight: SK,J,B(s)=bK,J,gJ,K,Jβ€²,|B|2(s).S_{K,J,B}(s)=b_{K,J,g_{J,K},J',|B|^2}(s).(8)✦

For a real function ww on ℝ\mathbb R, the integer periodization Per⁡w\operatorname{Per}w and folded circle density Fold⁡w\operatorname{Fold}w are the functions on ℝ\mathbb R and 𝕋\mathbb T, respectively, given by (Per⁡w)(t)=βˆ‘kβˆˆβ„€w(t+k),(Fold⁡w)(tβ€Ύ)=(Per⁡w)(t)(0<t≀1).(\operatorname{Per}w)(t)=\sum_{k\in\mathbb Z}w(t+k),\qquad (\operatorname{Fold}w)(\bar t)=(\operatorname{Per}w)(t)\quad(0<t\le1).(9) For a passage family with parameters (6), put K=[lβˆ’R,u+R]K=[l-R,u+R] and Ji(x)=miV(x)J_i(x)=m_iV(x). The unfolded marginal weight W𝒫,iW_{\mathcal P,i} of mode ii in a smooth passage family 𝒫\mathcal P is the real function on ℝ\mathbb R obtained by averaging the stationary density (8) over aβˆˆπ•‹a\in\mathbb T, as in (10).✦ The marginal density p𝒫,ip_{\mathcal P,i} of mode ii in a smooth passage family 𝒫\mathcal P is the real function on 𝕋\mathbb T obtained by folding the unfolded marginal weight W𝒫,iW_{\mathcal P,i} according to (9). The total marginal density p𝒫p_{\mathcal P} is the sum of these functions, and the physical packet sum F𝒫F_{\mathcal P} is the function Fm,AF_{m,A} on 𝕋3\mathbb T^3, as in (10). W𝒫,i(s)=βˆ«π•‹SK,Ji,A0,i(a,β‹…)(s)dΞΌ(a),p𝒫,i=Fold⁡W𝒫,i,p𝒫=βˆ‘i∈ιp𝒫,i,F𝒫=Fm,A.\begin{gathered} W_{\mathcal P,i}(s)=\int_{\mathbb T} S_{K,J_i,A_{0,i}(a,\cdot)}(s)\,d\mu(a),\qquad p_{\mathcal P,i}=\operatorname{Fold}W_{\mathcal P,i},\\ p_{\mathcal P}=\sum_{i\in\iota}p_{\mathcal P,i},\qquad F_{\mathcal P}=F_{m,A}. \end{gathered}(10)

The Claim

Theorem 1. Let Ο•:𝕋→𝕋\phi:\mathbb T\to\mathbb T be measurable, let mm be an injective finite list of integer modes, and let Ai,BiA_i,B_i be measurable square-integrable amplitudes on 𝕋2\mathbb T^2. Suppose that p,qp,q are spectral densities in the sense of (5) for σϕ,m,A,σϕ,m,B\sigma_{\phi,m,A},\sigma_{\phi,m,B}, respectively. Then βˆ«π•‹|pβˆ’q|dμ≀(βˆ₯Fm,Aβˆ₯2+βˆ₯Fm,Bβˆ₯2)d(Fm,A,Fm,B).\int_{\mathbb T}|p-q|\,d\mu \le\bigl(\|F_{m,A}\|_2+\|F_{m,B}\|_2\bigr) d(F_{m,A},F_{m,B}).(11)✦

Let 𝒫,𝒬\mathcal P,\mathcal Q be smooth passage families, possibly with different index sets, with the same speed VV. Assume that, within each family, at every slow point at most one amplitude with a given angular frequency is nonzero. For their total marginal densities and physical sums from (10), βˆ«π•‹|pπ’«βˆ’p𝒬|dμ≀(d(F𝒫,0)+d(F𝒬,0))d(F𝒫,F𝒬).\int_{\mathbb T}|p_{\mathcal P}-p_{\mathcal Q}|\,d\mu \le\bigl(d(F_{\mathcal P},0)+d(F_{\mathcal Q},0)\bigr) d(F_{\mathcal P},F_{\mathcal Q}).(12)✦

In the paper, the result reads: for two vectors F,GF,G with spectral densities pF,pGp_F,p_G for the same transformation, βˆ₯pFβˆ’pGβˆ₯1≀(βˆ₯Fβˆ₯2+βˆ₯Gβˆ₯2)βˆ₯Fβˆ’Gβˆ₯2\|p_F-p_G\|_1\le(\|F\|_2+\|G\|_2)\|F-G\|_2, where βˆ₯pFβˆ’pGβˆ₯1=βˆ«π•‹|pFβˆ’pG|dΞΌ\|p_F-p_G\|_1=\int_{\mathbb T}|p_F-p_G|\,d\mu.

What Is Checked

The comparison (11) is stated over finite packet spectral measures for a common measurable twist and injective mode list; (12) is stated over smooth passage families with a common speed and the stated restriction on coincident frequencies, using the physical distance (2). Finite packet sums (1) belong to L2(𝕋3,ΞΌ3)L^2(\mathbb T^3,\mu_3) whenever every amplitude belongs to L2(𝕋2,ΞΌ2)L^2(\mathbb T^2,\mu_2).✦