[Paper Review] General forms of the Menshov-Rademacher, Orlicz, and Tandori theorems on orthogonal series

Vladimir Mikhailets, Aleksandr Murach|arXiv (Cornell University)|Oct 19, 2011

Advanced Banach Space Theory9 references3 citations

TL;DR

This paper extends the classical Menshov–Rademacher, Orlicz, and Tandori theorems on orthogonal series convergence to the setting of direct integrals of measurable collections of Hilbert spaces, including vector-valued $L_2$-spaces $L_2(X,d au;H)$. It proves that the original convergence criteria—based on weighted $\ell^2$-norms involving $\log^2 n$ factors—remain valid for almost everywhere and unconditional convergence of orthogonal series in these generalized settings, with uniform bounds independent of the underlying measure space or Hilbert space dimension.

ABSTRACT

We prove that the classical Menshov-Rademacher, Orlicz, and Tandori theorems remain true for orthogonal series given in the direct integrals of measurable collections of Hilbert spaces. In particular, these theorems are true for the spaces L_{2}(X,dμ;H) of vector-valued functions, where (X,μ) is an arbitrary measure space, and H is a real or complex Hilbert space of an arbitrary dimension.

Motivation & Objective

• To extend the classical Menshov–Rademacher, Orlicz, and Tandori theorems on orthogonal series convergence to the broader setting of direct integrals of measurable collections of Hilbert spaces.
• To establish that the original convergence criteria—based on $\ell^2$-summability of coefficients weighted by $\log^2 n$—remain valid for orthogonal series in $L_2(X,d\mu;H)$, where $H$ is a Hilbert space of arbitrary dimension.
• To demonstrate that the convergence results hold for arbitrary measure spaces $X$, not just finite intervals with Lebesgue measure, and for both real and complex Hilbert spaces.
• To show that the theorems are stable under the direct integral construction, preserving the sharpness of the original conditions.
• To provide uniform bounds on the maximal function of partial sums, with explicit universal constants (e.g., $K=4$ in Theorem 1), independent of the underlying measure or Hilbert space structure.

Proposed method

• The authors work within the framework of direct integrals of Hilbert spaces $\mathbf{L}_2 = \int_X^\oplus H(x)\,d\mu(x)$, where $\{H(x)\}_{x\in X}$ is a $\mu$-measurable family of Hilbert spaces.
• They define the maximal function $S^*(\Phi,a,x) := \sup_m \|\sum_{n=1}^m a_n \varphi_n(x)\|_{H(x)}$ and establish $L^2$-bounds on this function via dyadic decomposition of the index set.
• The key technical tool is a dyadic decomposition of the sequence $ (a_n) $ into blocks $ M_k = \{n : \nu_k < n \leq \nu_{k+1}\} $ with $ \nu_k = 2^{2^k} $, enabling control of the $ \ell^2 $-norm of partial sums.
• The proofs rely on the Cauchy–Schwarz inequality applied to the $ L^2(X) $-norm of the maximal function, combined with estimates on the $ \ell^2 $-norms of coefficients over dyadic blocks.
• The authors use the fact that $ \sum_{k=0}^\infty \left( \sum_{n=\nu_k+1}^{\nu_{k+1}} |a_n|^2 \log^2_2 n \right)^{1/2} < \infty $ implies a.e. convergence, which is shown to be equivalent to the original Orlicz-type conditions.
• The extension to Riesz bases is established by modifying the constant in the maximal function estimate from $2 + \log_2 N$ to $C \log_2(N+1)$, with $C$ depending on the basis.

Experimental results

Research questions

• RQ1 Do the classical Menshov–Rademacher, Orlicz, and Tandori theorems on orthogonal series convergence remain valid for orthogonal series in direct integrals of Hilbert spaces?
• RQ2 Can the $ \ell^2(\log^2 n) $-type convergence criteria be extended to vector-valued functions with values in Hilbert spaces of arbitrary dimension?
• RQ3 Is the almost everywhere convergence of orthogonal series preserved when the underlying measure space is arbitrary (not necessarily finite or $\sigma$-finite) and the Hilbert spaces vary measurably over $X$?
• RQ4 Can the original sharpness of the theorems—i.e., the optimality of the $ \log^2 n $ weight—be preserved in the generalized setting?
• RQ5 Does the unconditional convergence result (Tandori theorem) extend to the direct integral framework under the same coefficient conditions as in the scalar case?

Key findings

• The classical Menshov–Rademacher theorem holds for orthogonal series in $ \mathbf{L}_2 = \int_X^\oplus H(x)\,d\mu(x) $: if $ \sum_{n=1}^\infty |a_n|^2 \log^2_2(n+1) < \infty $, then the series converges $ \mu $-a.e. on $ X $, with $ \|S^*(\Phi,a,\cdot)\|_2 \leq 4\sqrt{L} $.
• The Orlicz theorem extends to the direct integral setting: if $ \sum_{n=3}^\infty |a_n|^2 \log^2_2 n \cdot \omega_n < \infty $ for an increasing weight sequence $ (\omega_n) $ satisfying $ \sum_{n=2}^\infty \frac{1}{n (\log_2 n) \omega_n} < \infty $, then the series converges unconditionally $ \mu $-a.e.
• The Tandori theorem is generalized: under the same conditions as in the Orlicz theorem, the orthogonal series converges unconditionally $ \mu $-a.e., with the same coefficient condition ensuring convergence in the strong operator topology.
• The maximal function $ S^*(\Phi,a,x) $ is shown to be in $ L^2(X) $ under the $ \ell^2(\log^2 n) $-condition, with a uniform bound independent of the measure space or Hilbert space dimension.
• The results are stable under replacement of an orthonormal system by a Riesz basis: the convergence criteria remain valid with a modified constant $ C \log_2(N+1) $ in the maximal function estimate, where $ C $ depends on the basis.
• The proofs are robust to general measure spaces, including those that are not $ \sigma $-finite, by decomposing $ X $ into $ \sigma $-finite measurable subsets and applying the results locally.