OctoInformBeta

OpenAI preprint, September 2026 · The companion's direct model (no numbered result of this preprint)

The direct model with its ordinary and polar structure: Commuting operators SS, VV with a transitive algebra, SS-invariant coordinate tails, the polar decomposition S=U|S|S=U|S| and atomic and diffuse algebras of multipliers

One direct model has an ordinary and a polar structure.

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 ℤ2\mathbb{Z}_2 be the 2-adic integers, μ\mu the completion of the Haar measure on ℤ2\mathbb{Z}_2 of total mass one, and x↦x+1x \mapsto x+1 the odometer of ℤ2\mathbb{Z}_2. The space is ℋ=L2(ℤ2,μ;ℓ2(ℤ)),\mathcal{H} = L^2(\mathbb{Z}_2, \mu; \ell^2(\mathbb{Z})),(1)✦ the square-integrable functions ff from ℤ2\mathbb{Z}_2 to the fiber ℓ2(ℤ)\ell^2(\mathbb{Z}). We write f(x)jf(x)_j for the jjth coordinate of f(x)f(x) and eje_j for the jjth unit vector of ℓ2(ℤ)\ell^2(\mathbb{Z}).

The item concerns two operators on ℋ\mathcal{H} that act fiber by fiber: a forward operator SS, a weighted bilateral shift on each fiber, and a backward operator VV, which shifts the other way and carries the fiber at xx to the fiber at x+1x+1.

The Weights

The forward operator is weighted by numbers βx(j)\beta_x(j) read from the binary digits of x∈ℤ2x \in \mathbb{Z}_2. For x∈ℤ2x \in \mathbb{Z}_2, the plus residue sequence rxr_x of xx is the sequence of residues rx(m)∈{0,1,…,2m−1}r_x(m) \in \{0, 1, \dots, 2^m - 1\} of xx modulo 2m2^m, m≥0m \ge 0, and its minus residue sequence r‾x\bar r_x is the sequence given by r‾x(m)=2m−1−rx(m)\bar r_x(m) = 2^m - 1 - r_x(m). The weight block of an integer k≥1k \ge 1 is κ(k)=⌊log⁡8k⌋+1\kappa(k) = \lfloor \log_8 k \rfloor + 1✦, and its residue scale is 2κ(k)2^{\kappa(k)}.✦ For a sequence tt of natural numbers, the half exponent ψt\psi_t and its residue correction ctc_t are ψt(k)=10(k+1)+ct(k),ct(0)=0,ct(k)=100(2κ(k)−2+t(κ(k)))(k≥1).\psi_t(k) = 10(k+1) + c_t(k), \qquad c_t(0) = 0, \qquad c_t(k) = 100\bigl(2^{\kappa(k)} - 2 + t(\kappa(k))\bigr) \quad (k \ge 1).(2) For two such sequences tt and uu, the bilateral exponent φt,u\varphi_{t,u} on ℤ\mathbb{Z} is ψt(j)\psi_t(j) at j≥0j \ge 0 and ψu(−j−1)\psi_u(-j-1) at j<0j < 0✦, and the bilateral weight is e−φt,ue^{-\varphi_{t,u}}.✦ The weight of xx is the bilateral weight of its two residue sequences: βx(j)=e−φrx,r‾x(j),j∈ℤ.\beta_x(j) = e^{-\varphi_{r_x, \bar r_x}(j)}, \qquad j \in \mathbb{Z}.(3)✦

The Direct Model

For h∈ℤh \in \mathbb{Z}, the coordinate tail ℓ≥h2\ell^2_{\ge h} is the closed subspace of the v∈ℓ2(ℤ)v \in \ell^2(\mathbb{Z}) with vj=0v_j = 0 for every j<hj < h.✦ A subspace LL of a Hilbert space is invariant under an operator TT when Tv∈LTv \in L for every v∈Lv \in L.✦ Each g∈L∞(ℤ2,μ)g \in L^\infty(\mathbb{Z}_2, \mu) defines the multiplier MgM_g on ℋ\mathcal{H}, (Mgf)(x)=g(x)f(x)(M_g f)(x) = g(x) f(x).✦ An algebra 𝒜\mathcal{A} of operators on ℋ\mathcal{H} is transitive when the only closed subspaces of ℋ\mathcal{H} invariant under every member of 𝒜\mathcal{A} are 00 and ℋ\mathcal{H}.✦ For operators SS and VV on ℋ\mathcal{H}, 𝒜(S,V)\mathcal{A}(S, V) denotes the algebra generated by SS, VV and all the multipliers MgM_g.✦ The commutant {T}′\{T\}' of an operator TT on ℋ\mathcal{H} is the algebra of operators on ℋ\mathcal{H} that commute with TT.

A direct model on ℋ\mathcal{H} consists of operators SxS_x and VxV_x on ℓ2(ℤ)\ell^2(\mathbb{Z}) and real numbers bx(j)b_x(j), for x∈ℤ2x \in \mathbb{Z}_2 and j∈ℤj \in \mathbb{Z}, and operators SS and VV on ℋ\mathcal{H}, with the following properties. The functions x↦βx(j)x \mapsto \beta_x(j) and x↦bx(j)x \mapsto b_x(j) are measurable, βx(j)>0\beta_x(j) > 0 and 0≤bx(j)≤10 \le b_x(j) \le 1, and (Sxv)j=βx(j−1)vj−1,(Vxv)j=bx(j+1)vj+1(S_x v)_j = \beta_x(j-1)\, v_{j-1}, \qquad (V_x v)_j = b_x(j+1)\, v_{j+1}(4) for every xx, jj and v∈ℓ2(ℤ)v \in \ell^2(\mathbb{Z}). The operators SS and VV are assembled from these: for every f∈ℋf \in \mathcal{H} and almost every xx, (Sf)(x)=Sxf(x),(Vf)(x+1)=Vxf(x).(Sf)(x) = S_x f(x), \qquad (Vf)(x+1) = V_x f(x).(5) For every xx, the closed subspaces of ℓ2(ℤ)\ell^2(\mathbb{Z}) invariant under SxS_x are exactly 00, ℓ2(ℤ)\ell^2(\mathbb{Z}) and the coordinate tails. Moreover SV=VSSV = VS, ∥V∥≤1\|V\| \le 1, and ∥S∥=e−10,∥Sn∥≤exp⁡(−10⌊(n+1)2/4⌋)(n≥1).\|S\| = e^{-10}, \qquad \|S^n\| \le \exp\bigl(-10 \lfloor (n+1)^2/4 \rfloor\bigr) \quad (n \ge 1).(6) There is a measurable set X*⊆ℤ2X^* \subseteq \mathbb{Z}_2 of full measure, equal to its preimage under the odometer, such that for x∈X*x \in X^* every bx(j)b_x(j) is positive, VxV_x has dense range, and VxV_x maps each tail ℓ≥h2\ell^2_{\ge h} onto a dense subspace of ℓ≥h−12\ell^2_{\ge h-1}. The algebra 𝒜(S,V)\mathcal{A}(S, V) is transitive.

Projections and Strongly Closed Algebras

Let 𝒦\mathcal{K} be a complex Hilbert space and B(𝒦)B(\mathcal{K}) the algebra of bounded operators on 𝒦\mathcal{K}. The strong operator topology on B(𝒦)B(\mathcal{K}) is the topology induced by the maps A↦AxA \mapsto Ax, x∈𝒦x \in \mathcal{K}✦; a unital **-subalgebra of B(𝒦)B(\mathcal{K}) is strongly closed when it is closed in this topology.✦ For a set Σ\Sigma of operators on 𝒦\mathcal{K}, W*(Σ)W^*(\Sigma) is the intersection of all strongly closed unital **-subalgebras of B(𝒦)B(\mathcal{K}) that contain Σ\Sigma.✦

A minimal projection of a unital **-subalgebra ℳ\mathcal{M} of B(𝒦)B(\mathcal{K}) is a nonzero orthogonal projection P∈ℳP \in \mathcal{M} such that every orthogonal projection Q∈ℳQ \in \mathcal{M} with QP=QQP = Q is 00 or PP.✦ A unital **-subalgebra ℳ\mathcal{M} of B(𝒦)B(\mathcal{K}) is diffuse when it has no minimal projection✦, and atomic when it has a countable family of minimal projections PiP_i with PiPk=0(i≠k),∑iPiv=v(v∈𝒦),P_i P_k = 0 \quad (i \ne k), \qquad \sum_i P_i v = v \quad (v \in \mathcal{K}),(7)✦ the sum converging unconditionally. An operator TT on 𝒦\mathcal{K} has a pure point resolution when there is a countable family of nonzero orthogonal projections PiP_i satisfying (7) and there are real numbers λi\lambda_i with TPi=λiPiT P_i = \lambda_i P_i.✦

An operator TT on ℋ\mathcal{H} is a multiplier on ℤ2×ℤ\mathbb{Z}_2 \times \mathbb{Z} when there is g∈L∞(ℤ2×ℤ,μ×#)g \in L^\infty(\mathbb{Z}_2 \times \mathbb{Z}, \mu \times \#), with #\# the counting measure on ℤ\mathbb{Z}, such that (Tf)(x)j=g(x,j)f(x)j(Tf)(x)_j = g(x,j)\, f(x)_j for every f∈ℋf \in \mathcal{H}, almost every xx and every jj.✦

The Claim

Theorem 1. There is a direct model (Sx,S,b,Vx,V)(S_x, S, b, V_x, V) on ℋ\mathcal{H} with the following ordinary and polar structure.✦

Ordinary structure. For every h∈ℤh \in \mathbb{Z}, the subspace ℒh\mathcal{L}_h of the f∈ℋf \in \mathcal{H} with f(x)∈ℓ≥h2f(x) \in \ell^2_{\ge h} for almost every xx is closed, nonzero and proper, invariant under SS and not invariant under VV. Every fiber operator SxS_x is compact, and SS is not. The orthogonal projection QQ given by (Qf)(x)=f(x)0e0(Qf)(x) = f(x)_0\, e_0 commutes with every multiplier MgM_g, does not commute with SS, and is not MgM_g for any g∈L∞(ℤ2,μ)g \in L^\infty(\mathbb{Z}_2, \mu).

Polar structure. There is a polar family: a unitary UU on ℋ\mathcal{H} with (Uf)(x)j=f(x)j−1(Uf)(x)_j = f(x)_{j-1} and operators AnA_n on ℋ\mathcal{H}, n∈ℤn \in \mathbb{Z}, such that A0=|S|=(S*S)1/2,S=UA0,An=U−nA0Un,(Anf)(x)j=βx(j+n)f(x)j,A_0 = |S| = (S^*S)^{1/2}, \qquad S = U A_0, \qquad A_n = U^{-n} A_0 U^n, \qquad (A_n f)(x)_j = \beta_x(j+n)\, f(x)_j,(8)✦ the last for every ff, almost every xx and every jj. The operator A0A_0 is injective, has dense range and has a pure point resolution. For every finite set F⊂ℤF \subset \mathbb{Z} the algebra W*({An:n∈F})W^*(\{A_n : n \in F\}) is atomic. The algebra W*({An:n∈ℤ})W^*(\{A_n : n \in \mathbb{Z}\}) is abelian and diffuse, consists exactly of the multipliers on ℤ2×ℤ\mathbb{Z}_2 \times \mathbb{Z}, and is not contained in the commutant {S}′\{S\}'.

What Is Checked

The claim is stated over the space ℋ\mathcal{H} of (1), with the completed Haar measure μ\mu and the odometer of ℤ2\mathbb{Z}_2, the weights βx\beta_x of (3), the coordinate tails, the multipliers MgM_g and the multipliers on ℤ2×ℤ\mathbb{Z}_2 \times \mathbb{Z}, transitive algebras, the strong operator topology with the algebras W*(Σ)W^*(\Sigma), minimal projections, atomic and diffuse algebras, and pure point resolutions.

About these constructions the library proves that the self-adjoint operators on ℋ\mathcal{H} have a continuous functional calculus over ℝ\mathbb{R} for functions vanishing at 00✦, the calculus that defines |S|=(S*S)1/2|S| = (S^*S)^{1/2}, and that real scalars pass through products of operators on ℋ\mathcal{H}: t(AB)=(tA)B=A(tB)t(AB) = (tA)B = A(tB) for t∈ℝt \in \mathbb{R}.