OctoInformBeta

OpenAI preprint, September 2026 · §3. Modular localization and averaging

Lemma 3.2: The Fourier filter yg=∫g(t)σt(y)dty_g=\int g(t)\sigma_t(y)\,dt has norm at most ∥g∥1∥y∥\|g\|_1\|y\| and sends ξ\xi to f(D)yξf(D)y\xi for f(u)=∫g(t)eitudtf(u)=\int g(t)e^{itu}\,dt

Fourier filters along the modular flow stay in MM..

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

Let MM be a von Neumann algebra on a complex Hilbert space HH, ξ\xi a cyclic and separating unit vector for MM, Δ\Delta its modular operator and D=log⁡ΔD=\log\Delta. Averaging an element yy of MM along the modular automorphism group σt\sigma_t against an integrable kernel gg on ℝ\mathbb R gives an operator ygy_g, the Fourier filter of yy by gg. The question is whether ygy_g is again in MM, and how ygξy_g\xi comes from yξy\xi through the functional calculus of DD.

Standard Modular Data

A real spectral calculus DD on HH is a unital **-homomorphism f↦f(D)f\mapsto f(D) from the C*-algebra Cb(ℝ)C_b(\mathbb R) of bounded continuous complex functions on ℝ\mathbb R into B(H)B(H) that is nondegenerate: the vectors f(D)kf(D)k, with ff of compact support and k∈Hk\in H, form a dense subset of HH.✦

For a real spectral calculus DD on HH and real tt, the operator UtU_t is the image of the character et(u)=eitue_t(u)=e^{itu} of ℝ\mathbb R: Ut=et(D)=eitD.U_t=e_t(D)=e^{itD}.(1)✦

For a real spectral calculus DD on HH, the graph Γ\Gamma of eD/2e^{D/2} is the set of pairs (h,k)(h,k) in H×HH\times H such that fn(D)h→k(n→∞),fn(u)=emin⁡(n,u)/2.f_n(D)h\to k\quad(n\to\infty),\qquad f_n(u)=e^{\min(n,u)/2}.(2)✦ The function fnf_n is the cutoff of eu/2e^{u/2} at u=nu=n: it is bounded and continuous, positive, and bounded by en/2e^{n/2}.

Standard modular data (M,ξ,D,J)(M,\xi,D,J) on HH consist of a von Neumann algebra MM on HH, a unit vector ξ\xi cyclic and separating for MM, a real spectral calculus DD on HH and an antiunitary involution JJ of HH, such that {(aξ,a*ξ):a∈M}¯={(h,Jk):(h,k)∈Γ}.\overline{\{(a\xi,a^*\xi):a\in M\}}=\{(h,Jk):(h,k)\in\Gamma\}.(3)✦ Condition (3) is the polar decomposition S=JΔ1/2S=J\Delta^{1/2} of the closed Tomita operator SS, with modular operator Δ=eD\Delta=e^D and modular conjugation JJ; thus Ut=ΔitU_t=\Delta^{it}.

The modular flow of standard modular data is the family of maps σt\sigma_t on B(H)B(H), t∈ℝt\in\mathbb R, given on every bounded operator yy on HH by σt(y)=UtyU−t.\sigma_t(y)=U_t\,y\,U_{-t}.(4)✦

Strong Integrals Along the Flow

Let EE be a complex Banach space, gg an integrable complex function on ℝ\mathbb R with ∥g∥1=∫ℝ|g(t)|dt\|g\|_1=\int_{\mathbb R}|g(t)|\,dt, and t↦F(t)t\mapsto F(t) a family of bounded operators on EE with ∥F(t)∥≤C\|F(t)\|\le C for every tt and t↦F(t)ht\mapsto F(t)h continuous for every h∈Eh\in E. The strong integral ∫ℝg(t)F(t)dt\int_{\mathbb R}g(t)F(t)\,dt is the bounded operator on EE, of norm at most C∥g∥1C\|g\|_1, given by (∫ℝg(t)F(t)dt)h=∫ℝg(t)F(t)hdt,h∈E.\Bigl(\int_{\mathbb R}g(t)F(t)\,dt\Bigr)h=\int_{\mathbb R}g(t)\,F(t)h\,dt,\qquad h\in E.(5)✦

For standard modular data on HH, the Fourier filter ygy_g of a bounded operator yy on HH by an integrable function gg is the strong integral yg=∫ℝg(t)σt(y)dt.y_g=\int_{\mathbb R}g(t)\,\sigma_t(y)\,dt.(6)✦

The Claim

Theorem 1. Let gg be an integrable complex function on ℝ\mathbb R.

(a) Let (M,ξ,D,J)(M,\xi,D,J) be standard modular data on HH and yy a bounded operator on HH. Then ∥yg∥≤∥g∥1∥y∥,\|y_g\|\le\|g\|_1\,\|y\|,(7)✦ and if y∈My\in M, then yg∈M.y_g\in M.(8)✦

(b) Let ff be a bounded continuous function on ℝ\mathbb R with f(u)=∫ℝg(t)eitudtf(u)=\int_{\mathbb R}g(t)e^{itu}\,dt for every real uu. For standard modular data (M,ξ,D,J)(M,\xi,D,J) on HH and every bounded operator yy on HH, ygξ=f(D)yξ,y_g\xi=f(D)\,y\xi,(9)✦ and for every real spectral calculus DD on HH and every h∈Hh\in H, f(D)h=∫ℝg(t)Uthdt.f(D)h=\int_{\mathbb R}g(t)\,U_th\,dt.(10)✦

A vector hh has DD-spectral support in a set I⊂ℝI\subset\mathbb R when f(D)h=0f(D)h=0 for every bounded continuous ff vanishing on II, and compact DD-spectral support when II can be taken compact. In the paper, the result reads: for f∈Cc∞(ℝ)f\in C_c^\infty(\mathbb R) with f(u)=∫g(t)eitudtf(u)=\int g(t)e^{itu}\,dt and y∈My\in M, the weak integral yf=∫g(t)σt(y)dty_f=\int g(t)\sigma_t(y)\,dt, written ygy_g here, belongs to MM, with ∥yf∥≤∥g∥1∥y∥\|y_f\|\le\|g\|_1\|y\| and yfξ=f(D)yξy_f\xi=f(D)y\xi; if a∈Ma\in M and aξa\xi has compact DD-spectral support, there is a♭∈Ma^\flat\in M with aξ=ξa♭a\xi=\xi a^\flat, where ξa♭\xi a^\flat stands for J(a♭)*JξJ(a^\flat)^*J\xi; and the same construction applies to any continuous state-preserving action and its unitary implementation.

What Is Checked

The claim is stated over a complex Hilbert space HH, an integrable kernel gg and a bounded continuous function ff tied to it by f(u)=∫ℝg(t)eitudtf(u)=\int_{\mathbb R}g(t)e^{itu}\,dt, a real spectral calculus DD with its unitaries UtU_t, standard modular data (M,ξ,D,J)(M,\xi,D,J) with the graph of eD/2e^{D/2} in (3) and the modular flow σt\sigma_t, and the Fourier filter ygy_g, a strong integral in the sense of (5).

Beyond the claim, the library proves that for every bounded operator yy on HH and every h∈Hh\in H the map t↦σt(y)ht\mapsto\sigma_t(y)h from ℝ\mathbb R to HH is continuous.✦