By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

All papers

Claims

Findings

Backward intertwiners and a transitive commutant

Every infinite-dimensional separable complex Hilbert space admits a nonzero bounded norm-quasinilpotent operator with no nonzero proper closed hyperinvariant subspace, and its commutant is a proper strongly closed unital transitive complex operator algebra.

On H0=L2(Z2,μ;ℓ2(Z))H_0 = L^2(\mathbb{Z}_2, \mu; \ell^2(\mathbb{Z})), there exist measurable positive weights βj:Z2→(0,∞)\beta_j : \mathbb{Z}_2 \to (0, \infty) (j∈Zj \in \mathbb{Z}) and a contraction V∈B(H0)V \in \mathcal{B}(H_0) such that the shifts (Sf)(x)=Sxf(x)(Sf)(x) = S_x f(x) with Sxej=βj(x)ej+1S_x e_j = \beta_j(x) e_{j+1} define a nonzero bounded operator SS with ∥S∥=e−10\|S\| = e^{-10}, ∥Sn∥≤exp⁡(−10⌊(n+1)2/4⌋)\|S^n\| \le \exp(-10 \lfloor (n+1)^2/4 \rfloor) for all n≥1n \ge 1, every SxS_x has closed invariant subspaces precisely K≥hK_{\ge h} (h∈Z∪{−∞,+∞}h \in \mathbb{Z} \cup \{-\infty, +\infty\}), VV commutes with SS and has the form (Vf)(θx)=Vxf(x)(Vf)(\theta x) = V_x f(x) with Vxej=bj(x)ej−1V_x e_j = b_j(x) e_{j-1} (0≤bj(x)≤10 \le b_j(x) \le 1 positive on a common conull set where VxK≥h‾=K≥h−1\overline{V_x K_{\ge h}} = K_{\ge h-1} and VxK‾=K\overline{V_x K} = K), and the unital complex algebra generated by SS, VV, and all scalar multipliers MgM_g is transitive.

If w=(wj)j∈Zw = (w_j)_{j \in \mathbb{Z}} is positive and decreasing, log⁡wj\log w_j is convex for j≤0j \le 0 and concave for j≥0j \ge 0, ∑j≠0wj1/j<∞\sum_{j \ne 0} w_j^{1/j} < \infty, and lim⁡n→∞log⁡wnn2=−a\lim_{n \to \infty} \frac{\log w_n}{n^2} = -a and lim⁡n→∞log⁡w−nn2=a\lim_{n \to \infty} \frac{\log w_{-n}}{n^2} = a for some a>log⁡3a > \log 3, then the closed RR-invariant subspaces of ℓ2(w)\ell^2(w) are precisely {z:zj=0 for j<h}\{z : z_j = 0 \text{ for } j < h\} for h∈Z∪{−∞,+∞}h \in \mathbb{Z} \cup \{-\infty, +\infty\}, and the shift Tej=(wj+1/wj)ej+1T e_j = (w_{j+1}/w_j) e_{j+1} on KK has invariant lattice {K≥h:h∈Z∪{−∞,+∞}}\{K_{\ge h} : h \in \mathbb{Z} \cup \{-\infty, +\infty\}\}.

For every x∈X=Z2x \in X = \mathbb{Z}_2, the operator Sxej=βj(x)ej+1S_x e_j = \beta_j(x) e_{j+1} is bounded, has norm e−10e^{-10}, has precisely the standard tails K≥hK_{\ge h} as its closed invariant subspaces, and x↦Sxvx \mapsto S_x v is measurable for each v∈Kv \in K.

The operator SS on H0H_0 is nonzero and norm-quasinilpotent, satisfying ∥Sn∥≤exp⁡(−10⌊(n+1)24⌋)\|S^n\| \le \exp\left(-10 \left\lfloor \frac{(n+1)^2}{4} \right\rfloor\right) for all integers n≥1n \ge 1.

There is a contraction V∈B(H0)V \in \mathcal{B}(H_0) of the form (Vf)(θx)=Vxf(x)(Vf)(\theta x) = V_x f(x) with Vxej=bj(x)ej−1V_x e_j = b_j(x) e_{j-1} (0≤bj(x)≤10 \le b_j(x) \le 1 positive on a θ\theta-invariant conull set X∗X_*) such that VV commutes with SS and VxK≥h‾=K≥h−1\overline{V_x K_{\ge h}} = K_{\ge h-1} (h∈Zh \in \mathbb{Z}) and VxK‾=K\overline{V_x K} = K for every x∈X∗x \in X_*.

If θ\theta is an invertible ergodic probability-preserving transformation of (Y,ν)(Y, \nu) and a measurable function s:Y→Z∪{−∞,+∞}s : Y \to \mathbb{Z} \cup \{-\infty, +\infty\} satisfies almost everywhere s(x)=−∞  ⟹  s(θx)=−∞s(x) = -\infty \implies s(\theta x) = -\infty and s(x)∈Z  ⟹  s(θx)≤s(x)−1s(x) \in \mathbb{Z} \implies s(\theta x) \le s(x) - 1, then s=−∞s = -\infty almost everywhere or s=+∞s = +\infty almost everywhere.

For the polar-conjugate family An=U−n∣S∣UnA_n = U^{-n}|S|U^n (n∈Zn \in \mathbb{Z}), the von Neumann algebra W∗(An:n∈F)W^*(A_n : n \in F) is atomic for every finite F⊂ZF \subset \mathbb{Z}, whereas W∗(An:n∈Z)=L∞(X×Z,μ×#)W^*(A_n : n \in \mathbb{Z}) = L^\infty(X \times \mathbb{Z}, \mu \times \#) is diffuse.

Methods

Constructing a bounded bilateral weighted-shift field SS on L2(Z2,μ;ℓ2(Z))L^2(\mathbb{Z}_2, \mu; \ell^2(\mathbb{Z})) with 2-adic residue-dependent weights βj(x)=e−aj(x)\beta_j(x) = e^{-a_j(x)} whose logarithms grow quadratically with opposite signs at the two ends and shift in opposite directions under θx=x+1\theta x = x + 1.

Source excerpts

“a+ k (x) = c(k + 1) + Jm + DTm(x), (17) a− k (x) = c(k + 1) + Jm + D(Nm − 1 − Tm(x)). (18) On the full integer axis define aj (x) = { a+ j (x), j ≥ 0, a− −j−1 (x), j < 0, βj (x) = e −aj (x) . (19)” · Show in paper

Constructing a bounded backward intertwiner VV commuting with SS by solving the commutation recurrence b~j+1(x)=b~j(x)βj−1(θx)βj(x)\tilde{b}_{j+1}(x) = \tilde{b}_j(x) \frac{\beta_{j-1}(\theta x)}{\beta_j(x)} in both directions and normalizing each fiber by L(x)=sup⁡j∈Zb~j(x)L(x) = \sup_{j \in \mathbb{Z}} \tilde{b}_j(x).

Source excerpts

“For every x ∈ X set ̃b0(x) = 1 and define positive coefficients at all integer indices by ̃bj+1(x) = ̃bj (x) βj−1(θx) βj (x) (j ∈ Z). (26)” · Show in paper

“The countable supremum L(x) = sup j∈Z ̃bj (x) is measurable, and 1 ≤ L(x) < ∞ on X∗. Define bj (x) = { ̃bj (x)/L(x), x ∈ X∗, 0, x /∈ X∗.” · Show in paper

Classifying closed invariant subspaces of bilateral weighted translations via a discrete convex-envelope and support-separation argument.

Source excerpts

“The convex-envelope and support-separation argument below follows the method of Domar’s weighted-translation work [3, Section 3]. We prove the stated quadratic specialization in full, including the normalized estimates for both convolution tails.” · Show in paper

Proving transitivity of the algebra generated by SS, VV, and scalar multipliers MgM_g by reducing an invariant subspace to a measurable field of orthogonal projections PxP_x, encoding fiber invariant tails as a measurable label s(x)s(x), and applying an ergodic strict-descent argument.

Source excerpts

“The field lemma [9, Lemma 3.1], applied with the probability space (X, µ) and K = ℓ 2(Z), gives (P f)(x) = Pxf(x), M = {f ∈ H0 : f(x) ∈ ran Px for almost every x}” · Show in paper

“By Proposition 3.2 and (31), there is a label s(x) ∈ Z ∪ {−∞, +∞}, ran Px = K≥s(x).” · Show in paper