OctoInformBeta

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

The direct model: A nonzero fiberwise weighted shift SS and a contraction VV commuting with it, which with the L∞L^\infty multipliers generate a transitive algebra

An explicit direct model over the 2-adic odometer.

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 ring of 2-adic integers, μ\mu the completion of its Haar probability measure and τ(x)=x+1\tau(x) = x + 1 the odometer; measurable means μ\mu-measurable. The Hilbert space of the item is the space of square-integrable ℓ2(ℤ)\ell^2(\mathbb{Z})-valued functions on ℤ2\mathbb{Z}_2, ℋ=L2(ℤ2,μ;ℓ2(ℤ)),\mathcal{H} = L^2\bigl(\mathbb{Z}_2, \mu; \ell^2(\mathbb{Z})\bigr),(1)✦ and (ej)j∈ℤ(e_j)_{j \in \mathbb{Z}} is the standard basis of the fiber ℓ2(ℤ)\ell^2(\mathbb{Z}).

Two operators act on ℋ\mathcal{H}, both built from weighted shifts with weights read from the 2-adic digits of xx: a forward operator SS, which acts on each fiber separately, and a backward operator VV, which shifts the other way while carrying the fiber at xx to the fiber at x+1x + 1. The item concerns the invariant subspaces of the fiber shifts and of the algebra generated by SS, VV and the multiplications by bounded functions of xx, together with the norms of VV and of the powers of SS. The library gathers these properties into the notion of a direct model.

The Weights

For m≥0m \ge 0 and x∈ℤ2x \in \mathbb{Z}_2, the residue rm(x)r_m(x) of xx modulo 2m2^m is the integer in {0,1,…,2m−1}\{0, 1, \dots, 2^m - 1\} congruent to xx modulo 2m2^m. The plus and minus residues of xx are the sequences px(m)=rm(x),qx(m)=2m−1−rm(x),m≥0.p_x(m) = r_m(x), \qquad q_x(m) = 2^m - 1 - r_m(x), \qquad m \ge 0.(2)

The block of an index k≥1k \ge 1 is ν(k)=⌊log⁡8k⌋+1\nu(k) = \lfloor \log_8 k \rfloor + 1, with ν(0)=1\nu(0) = 1✦, and its scale is 2ν(k)2^{\nu(k)}✦. For a sequence tt of natural numbers, the residue correction κt\kappa_t and the half exponent ata_t are the sequences κt(0)=0,κt(k)=100(2ν(k)−2+t(ν(k)))(k≥1),at(k)=10(k+1)+κt(k),\kappa_t(0) = 0, \qquad \kappa_t(k) = 100\bigl(2^{\nu(k)} - 2 + t(\nu(k))\bigr) \ (k \ge 1), \qquad a_t(k) = 10(k+1) + \kappa_t(k),(3) and the partial sums of the half exponents are At(n)=∑k<nat(k)A_t(n) = \sum_{k < n} a_t(k)✦.

For two such sequences tt and uu, the bilateral exponent εt,u\varepsilon_{t,u} and the bilateral weight sequence wt,uw_{t,u} are the functions on ℤ\mathbb{Z} given by εt,u(j)={at(j),j≥0,au(−j−1),j<0,wt,u(j)={e−At(j),j≥0,eAu(−j),j<0,\varepsilon_{t,u}(j) = \begin{cases} a_t(j), & j \ge 0, \\ a_u(-j-1), & j < 0, \end{cases} \qquad w_{t,u}(j) = \begin{cases} e^{-A_t(j)}, & j \ge 0, \\ e^{A_u(-j)}, & j < 0, \end{cases}(4) and the bilateral weight is βt,u(j)=e−εt,u(j)\beta_{t,u}(j) = e^{-\varepsilon_{t,u}(j)}✦. The explicit weights and the explicit weight sequence of x∈ℤ2x \in \mathbb{Z}_2 are those of its plus and minus residues: βx=βpx,qx\beta_x = \beta_{p_x, q_x}✦ and wx=wpx,qxw_x = w_{p_x, q_x}✦.

The backward ratios of xx and their supremum are ρx(j)=wx+1(j−1)e10wx(j),Mx=supj∈ℤρx(j),\rho_x(j) = \frac{w_{x+1}(j-1)}{e^{10}\, w_x(j)}, \qquad M_x = \sup_{j \in \mathbb{Z}} \rho_x(j),(5) and the backward coefficients bx(j)b_x(j) of x∈ℤ2x \in \mathbb{Z}_2 and j∈ℤj \in \mathbb{Z} are the numbers in [0,1][0, 1] determined by the backward ratios ρx(j)\rho_x(j).

The Operators

For a sequence aa on ℤ\mathbb{Z} with |a(j)|≤C|a(j)| \le C for every jj, the diagonal operator DaD_a on ℓ2(ℤ)\ell^2(\mathbb{Z}) multiplies the coordinates by aa✦; for a bijection σ\sigma of ℤ\mathbb{Z}, the reindexing UσU_\sigma is the isometry of ℓ2(ℤ)\ell^2(\mathbb{Z}) onto itself that composes with σ\sigma: (Dav)(j)=a(j)v(j),(Uσv)(j)=v(σ(j)).(D_a v)(j) = a(j)\, v(j), \qquad (U_\sigma v)(j) = v(\sigma(j)).(6)✦

For a bijection σ\sigma of ℤ\mathbb{Z} and a bounded sequence aa on ℤ\mathbb{Z}, the weighted permutation operator of σ\sigma and aa is the bounded operator UσDaU_\sigma D_a on ℓ2(ℤ)\ell^2(\mathbb{Z}).✦ The forward fiber operator SxS_x is the weighted permutation operator of σ(j)=j−1\sigma(j) = j - 1 and the weights βx\beta_x, the weighted bilateral shift ej↦βx(j)ej+1e_j \mapsto \beta_x(j) e_{j+1}.✦ The backward fiber operator VxV_x is the weighted permutation operator of σ(j)=j+1\sigma(j) = j + 1 and the coefficients bxb_x, the backward weighted shift ej↦bx(j)ej−1e_j \mapsto b_x(j) e_{j-1}.✦

Let (X,ν)(X, \nu) be a measure space. The decomposable operator AA of a field x↦A(x)x \mapsto A(x) of operators on ℓ2(ℤ)\ell^2(\mathbb{Z}), with ∥A(x)∥≤C\|A(x)\| \le C for every xx and x↦A(x)f(x)x \mapsto A(x) f(x) measurable whenever ff is, and the multiplier MgM_g of a function g∈L∞(X,ν)g \in L^\infty(X, \nu) are the operators on L2(X,ν;ℓ2(ℤ))L^2(X, \nu; \ell^2(\mathbb{Z})) given by (Af)(x)=A(x)f(x),(Mgf)(x)=g(x)f(x).(A f)(x) = A(x) f(x), \qquad (M_g f)(x) = g(x) f(x).(7)

The forward operator SS is the decomposable operator of x↦Sxx \mapsto S_x on ℋ\mathcal{H}✦, and the backward operator VV is the decomposable operator of x↦Vxx \mapsto V_x followed by composition with the inverse odometer τ−1(x)=x−1\tau^{-1}(x) = x - 1: (Sf)(x)=Sxf(x),(Vf)(x)=Vx−1f(x−1).(S f)(x) = S_x f(x), \qquad (V f)(x) = V_{x-1} f(x - 1).(8)✦

The algebra 𝒜(T,W)\mathcal{A}(T, W) of two bounded operators TT and WW on ℋ\mathcal{H} is the complex subalgebra of B(ℋ)B(\mathcal{H}) generated by TT, WW and the multipliers MgM_g, g∈L∞(ℤ2,μ)g \in L^\infty(\mathbb{Z}_2, \mu).✦

A subspace LL of a complex normed space is invariant under a bounded operator TT when Tv∈LTv \in L for every v∈Lv \in L, and Lat⁡T\operatorname{Lat} T is the set of closed subspaces invariant under TT.✦ A subalgebra 𝒜\mathcal{A} of B(ℋ)B(\mathcal{H}) is transitive when every closed subspace of ℋ\mathcal{H} invariant under every member of 𝒜\mathcal{A} is 00 or ℋ\mathcal{H}.✦ For h∈ℤh \in \mathbb{Z}, the coordinate tail ℓ≥h2\ell^2_{\ge h} is the closed subspace of ℓ2(ℤ)\ell^2(\mathbb{Z}) of the vectors vv with v(j)=0v(j) = 0 for every j<hj < h.✦

Direct Models

A direct model is a quintuple ((Tx),T,c,(Wx),W)\bigl((T_x), T, c, (W_x), W\bigr) of a field (Tx)x∈ℤ2(T_x)_{x \in \mathbb{Z}_2} of bounded operators on ℓ2(ℤ)\ell^2(\mathbb{Z}), a bounded operator TT on ℋ\mathcal{H}, a real function cx(j)c_x(j) of x∈ℤ2x \in \mathbb{Z}_2 and j∈ℤj \in \mathbb{Z}, a field (Wx)x∈ℤ2(W_x)_{x \in \mathbb{Z}_2} of bounded operators on ℓ2(ℤ)\ell^2(\mathbb{Z}) and a bounded operator WW on ℋ\mathcal{H}, subject to the conditions below.✦ Write ℳ\mathcal{M} for the set of direct models.

The operators TxT_x are weighted shifts with the explicit weights βx(j)\beta_x(j), which are positive and measurable in xx, and TT is decomposable with fibers TxT_x: for all x∈ℤ2x \in \mathbb{Z}_2, j∈ℤj \in \mathbb{Z}, v∈ℓ2(ℤ)v \in \ell^2(\mathbb{Z}) and f∈ℋf \in \mathcal{H}, (Txv)(j)=βx(j−1)v(j−1),(Tf)(x)=Txf(x)for almost every x,(T_x v)(j) = \beta_x(j-1)\, v(j-1), \qquad (T f)(x) = T_x f(x) \ \text{for almost every } x,(9) and, for every x∈ℤ2x \in \mathbb{Z}_2, Lat⁡Tx={0,ℓ2(ℤ)}∪{ℓ≥h2:h∈ℤ}.\operatorname{Lat} T_x = \{0, \ell^2(\mathbb{Z})\} \cup \{\ell^2_{\ge h} : h \in \mathbb{Z}\}.(10)

The coefficients cx(j)c_x(j) lie in [0,1][0, 1] and are measurable in xx, and WW carries the fiber at xx to the fiber at x+1x + 1 by WxW_x: for all xx, jj, vv and ff, (Wxv)(j)=cx(j+1)v(j+1),(Wf)(x+1)=Wxf(x)for almost every x.(W_x v)(j) = c_x(j+1)\, v(j+1), \qquad (W f)(x + 1) = W_x f(x) \ \text{for almost every } x.(11)

The pair satisfies T≠0,∥T∥=e−10,∥Tn∥≤e−10⌊(n+1)2/4⌋(n≥1),∥W∥≤1,TW=WT,T \ne 0, \quad \|T\| = e^{-10}, \quad \|T^n\| \le e^{-10 \lfloor (n+1)^2/4 \rfloor} \ (n \ge 1), \quad \|W\| \le 1, \quad TW = WT,(12) and the algebra 𝒜(T,W)\mathcal{A}(T, W) is transitive.

Finally, there is a measurable set X*⊆ℤ2X^* \subseteq \mathbb{Z}_2 with μ(ℤ2\X*)=0\mu(\mathbb{Z}_2 \setminus X^*) = 0 and τ−1(X*)=X*\tau^{-1}(X^*) = X^* such that for every x∈X*x \in X^* the coefficients cx(j)c_x(j) are positive, WxW_x has dense range, and for every h∈ℤh \in \mathbb{Z} Wxℓ≥h2¯=ℓ≥h−12.\overline{W_x \ell^2_{\ge h}} = \ell^2_{\ge h-1}.(13)

The Claim

Theorem 1. There is a direct model: ℳ≠∅.\mathcal{M} \ne \emptyset.(14)✦ The explicit model, the quintuple of the forward fiber operators SxS_x, the forward operator SS, the backward coefficients bx(j)b_x(j), the backward fiber operators VxV_x and the backward operator VV, is one: ((Sx),S,b,(Vx),V)∈ℳ.\bigl((S_x), S, b, (V_x), V\bigr) \in \mathcal{M}.(15)✦

What Is Checked

The claim is stated over the space ℋ\mathcal{H} of ℓ2(ℤ)\ell^2(\mathbb{Z})-valued functions square-integrable for the completed Haar probability measure μ\mu on ℤ2\mathbb{Z}_2, the odometer τ\tau, the explicit weights βx(j)\beta_x(j) and coefficients bx(j)b_x(j) built from the plus and minus residues of xx, the weighted permutation operators SxS_x and VxV_x, the decomposable operators and L∞L^\infty multipliers on ℋ\mathcal{H}, the coordinate tails, and the algebra 𝒜(T,W)\mathcal{A}(T, W) with its transitivity.

About these constructions the library proves that for |a(j)|≤C|a(j)| \le C the sequence (a(j)v(j))j(a(j) v(j))_j lies in ℓ2(ℤ)\ell^2(\mathbb{Z})✦ with ∥Dav∥≤C∥v∥\|D_a v\| \le C \|v\|✦, that reindexing preserves the norm✦, and that the decomposable operator of a field bounded by CC satisfies ∥Af∥≤C∥f∥\|A f\| \le C \|f\|✦. It proves ∥Sx∥≤e−10\|S_x\| \le e^{-10}✦ and ∥Vx∥≤1\|V_x\| \le 1✦ for every xx, that x↦Sxf(x)x \mapsto S_x f(x)✦ and x↦Vxf(x)x \mapsto V_x f(x)✦ are measurable whenever ff is, and that the inverse odometer preserves μ\mu✦. It also proves that the multiplier MgM_g of g∈L∞(ℤ2,μ)g \in L^\infty(\mathbb{Z}_2, \mu) satisfies ∥Mgf∥≤∥g∥∞∥f∥\|M_g f\| \le \|g\|_\infty \|f\| for every f∈ℋf \in \mathcal{H}.