[Paper Review] Some maximal inequalities on Triebel-Lizorkin spaces for $p=\infty$

Bae Jun Park|arXiv (Cornell University)|Oct 3, 2017

Advanced Harmonic Analysis Research15 references3 citations

TL;DR

This paper establishes new maximal inequalities for Triebel-Lizorkin spaces at the endpoint $ p = ty $, introducing a modified maximal operator $ \mathcal{M}_r^{k,\epsilon} $ that overcomes the failure of classical Fefferman-Stein and Peetre-type inequalities in this regime. The key contribution is a sharp $ \dot{F}_\infty^{s,q} $-variant maximal inequality that enables boundedness in $ \ell^q $-sums over dyadic cubes, with applications to function space theory and sharpness results for the underlying conditions on exponents.

ABSTRACT

In this work we give some maximal inequalities in Triebel-Lizorkin spaces, which are "$\dot{F}_{\infty}^{s,q}$-variants" of Fefferman-Stein vector-valued maximal inequality and Peetre's maximal inequality. We will give some applications of the new maximal inequalities and discuss sharpness of some results.

Motivation & Objective

• To address the failure of classical Fefferman-Stein and Peetre maximal inequalities in the endpoint case $ p = \infty $ for Triebel-Lizorkin spaces $ \dot{F}_\infty^{s,q} $.
• To develop a new maximal operator $ \mathcal{M}_r^{k,\epsilon} $ that adapts to the structure of $ \dot{F}_\infty^{s,q} $ spaces and restores boundedness in $ \ell^q $-sums over dyadic cubes.
• To prove sharpness of the exponent condition $ r < q $ in the new maximal inequality, showing the condition is necessary.
• To provide applications of the new inequalities to function space theory, particularly in the context of $ BMO $ and $ bmo $ spaces.
• To extend the theory of maximal operators to the limiting case $ p = \infty $, where standard tools fail due to lack of $ L^p $-boundedness.

Proposed method

• Introduce a modified maximal operator $ \mathcal{M}_r^{k,\epsilon}f(x) $ that combines local $ L^r $ averages over dyadic cubes with a decay factor $ (2^k l(Q))^{-\epsilon} $ for large cubes.
• Use the frequency localization $ f_k \in \mathcal{E}(2^k) $ to control the behavior of $ \mathfrak{M}_{\sigma,2^k}f_k $ via $ \mathcal{M}_t^{k,\epsilon}f_k $, establishing a pointwise comparison via Lemma 1.2.
• Prove the main maximal inequality: for $ 0 < r < q < \infty $, the $ \ell^q $-sum of $ \mathcal{M}_r^{k,\epsilon}f_k $ over dyadic cubes is controlled by the $ \ell^q $-sum of $ |f_k| $ in the weak-type $ \dot{F}_\infty^{0,q} $-norm.
• Construct counterexamples using oscillatory functions $ f_k $ supported on shifted cubes to show that the classical $ \mathcal{M}_r $ fails for $ p = \infty $, even when $ r < q $.
• Use random construction techniques from Christ and Seeger to prove sharpness of the condition $ r < q $, showing that the inequality fails otherwise.
• Apply the new inequality to derive boundedness in $ \dot{F}_\infty^{s,q} $-norms and relate it to known results in $ BMO $ and $ bmo $ spaces.

Experimental results

Research questions

• RQ1 Can the Fefferman-Stein vector-valued maximal inequality be extended to the endpoint case $ p = \infty $ in Triebel-Lizorkin spaces?
• RQ2 Why do classical maximal operators fail in the $ \dot{F}_\infty^{s,q} $-setting, and what structural modifications are needed to restore boundedness?
• RQ3 Is the condition $ r < q $ necessary for the maximal inequality to hold in the $ \dot{F}_\infty^{s,q} $-framework?
• RQ4 Can a new maximal operator be constructed that respects the dyadic and frequency-localized structure of $ \dot{F}_\infty^{s,q} $ spaces?
• RQ5 What are the sharp conditions on the exponents $ r, q $ for the boundedness of the maximal operator in the $ \ell^q $-sum over dyadic cubes?

Key findings

• The classical Fefferman-Stein inequality fails for $ p = \infty $, even when $ r < q $, as shown by a counterexample with $ f_k $ supported on shifted cubes.
• The modified maximal operator $ \mathcal{M}_r^{k,\epsilon} $ satisfies the key inequality: $ \left\| \left( \sum_k (\mathcal{M}_r^{k,\epsilon} f_k)^q \right)^{1/q} \right\|_{\dot{F}_\infty^{0,q}} \lesssim \left\| \left( \sum_k |f_k|^q \right)^{1/q} \right\|_{\dot{F}_\infty^{0,q}} $ for $ 0 < r < q < \infty $.
• The condition $ r < q $ is sharp: if $ r \geq q $, the inequality fails, as demonstrated via a random construction and the use of $ \mathfrak{M}_{\sigma,2^k} $ operators.
• For $ f_k \in \mathcal{E}(2^k) $, the operator $ \mathfrak{M}_{d/r,2^k}f_k $ is pointwise controlled by $ \mathcal{M}_t^{k,d(1/r-1/t)}f_k $ for $ r < t $, enabling the use of $ \mathcal{M}_r^{k,\epsilon} $ in place of $ \mathcal{M}_r $.
• The inequality holds uniformly over dyadic cubes $ P $, with the norm defined via $ \sup_P \left( \frac{1}{|P|} \int_P \sum_k |f_k(x)|^q dx \right)^{1/q} $, which characterizes $ \|f\|_{\dot{F}_\infty^{0,q}} $.
• The sharpness result shows that $ \sup_P \left( \frac{1}{|P|} \int_P \sum_k (\mathfrak{M}_{\sigma,2^k}f_k^\mu)^q dx \right)^{1/q} \gtrsim \max\{2^{N(d/q - \sigma)}, N^{1/q}\} $ for $ \sigma \leq d/q $, proving necessity of $ r < q $.