Function spaces and potential theory in the Orlicz setting

1. Problem Statement and Motivation

The paper addresses the need to extend classical potential theory and function space theory beyond the standard $L^p$ (power-law) framework to the more general setting of Orlicz spaces.

• Context: Classical potential theory relies heavily on Bessel and Riesz potentials defined via convolution with kernels like $G_s$ (Bessel) and $I_s$ (Riesz). In the $L^p$ setting, these spaces (Bessel potential spaces $H^{s,p}$, Sobolev spaces $W^{s,p}$, and Lizorkin-Triebel spaces $F^s_{p,q}$) are well-understood, with established equivalences (e.g., $H^{k,p} = W^{k,p}$ for integer $k$) and embedding theorems.
• Limitation: Many physical and geometric problems involve growth conditions that are not polynomial (e.g., exponential growth, logarithmic corrections). Standard $L^p$ spaces cannot capture these behaviors.
• Objective: The authors aim to define and analyze Bessel-Orlicz potential spaces ($H^{s,A}$) and Orlicz-Lizorkin-Triebel spaces ($F^{s,A,q}$) where the underlying integrability is governed by a general Young function $A$ rather than a power $t^p$. They seek to:
  1. Establish the equivalence between potential-based definitions and derivative-based definitions (Orlicz-Sobolev spaces).
  2. Prove inclusion results for fractional orders.
  3. Derive pointwise decay estimates (Strauss-type lemmas) for radial functions.
  4. Provide atomic decompositions for these spaces.

2. Methodology

The authors employ a blend of harmonic analysis, singular integral theory, and Orlicz space techniques.

• Definitions:

  • Bessel-Orlicz Space ($H^{s,A}$): Defined as the set of functions $u = G_s * f$ where $f \in L^A(\mathbb{R}^n)$, with the norm $\|u\|_{H^{s,A}} = \|f\|_{L^A}$. Here, $G_s$ is the Bessel kernel, the inverse Fourier transform of $(1+|\xi|^2)^{-s/2}$.
  • Orlicz-Sobolev Space ($W^{s,A}$): Defined via the modular involving the Young function $A$ applied to the function and its fractional difference quotient (or gradient for $s=1$).
  • Orlicz-Lizorkin-Triebel Space ($F^{s,A,q}$): Defined via Littlewood-Paley decomposition using a dyadic partition of unity, replacing $L^p$ norms with Orlicz norms.

• Key Technical Tools:

  • Singular Integral Operators: A major challenge is that the gradient of the Bessel kernel, $\nabla G_1$, is not in $L^1(\mathbb{R}^n)$. The authors prove that $\nabla G_1$ defines a Calderón-Zygmund singular integral operator. They utilize the boundedness of these operators on Orlicz spaces (under $\Delta_2$ conditions) to bridge the gap between potential and derivative definitions.
  • Continuity of Bessel Kernels: They establish a novel continuity result (Proposition 3.8) for the kernel $K(z, x, t) = (G_\alpha(z-x) - G_\alpha(z-x+t))|t|^{-\gamma}$ in mixed Orlicz spaces. This avoids reliance on the homogeneity of power functions, which is a standard assumption in classical proofs.
  • Radial Analysis: For the Strauss lemma, they exploit the symmetry of radial functions and the decay properties of the Bessel kernel to derive pointwise bounds.
  • Atomic Decomposition: They adapt the classical dyadic decomposition techniques (using $C^m$-atoms on dyadic cubes) to the Orlicz setting, utilizing the Fefferman-Stein maximal theorem for Orlicz spaces.

3. Key Contributions and Results

A. Equivalence of Spaces (Integer Order)

Theorem 3.4 & 3.5: Under the assumption that the Young function $A$ and its conjugate $\hat{A}$ satisfy the $\Delta_2$ condition (growth control), the authors prove:
$$H^{1,A}(\mathbb{R}^n) = W^{1,A}(\mathbb{R}^n)$$
This is generalized to any integer order $m \in \mathbb{N}$:
$$H^{m,A}(\mathbb{R}^n) = W^{m,A}(\mathbb{R}^n)$$
Significance: This recovers the classical Calderón-type theorem in the non-standard Orlicz setting, validating that the potential definition (convolution with Bessel kernel) is equivalent to the derivative definition.

B. Fractional Order Inclusions

Theorem 3.7 & 3.9: The authors establish continuous embeddings between spaces of different fractional orders:

1. $H^{s',A} \subset W^{s,A}$ for $0 < s < s' < 1$. This holds for any Young function $A$, regardless of $\Delta_2$ conditions.
2. $W^{s',A} \subset H^{s,A}$ for $0 < s < s' < 1$. This requires $A$ to satisfy the $\Delta_2$ condition.
   Significance: These results generalize the classical Sobolev embedding theorems to the Orlicz framework, providing precise norm and modular relations.

C. Strauss-Type Lemma for Radial Functions

Theorem 4.3: For radial functions $u \in H^{s,A}_{rad}(\mathbb{R}^n)$ with $sp_- > 1$, the authors prove a pointwise decay estimate:
$$|u(x)| \leq C \frac{|x|^{1-n}}{\hat{A}^{-1}(|x|^{1-n})} \|u\|_{H^{s,A}}$$
Significance: This generalizes the classical Strauss inequality (which gives decay like $|x|^{-(n-1)/p}$ for $L^p$). It is crucial for the study of existence and regularity of solutions to non-local equations (e.g., Hénon-type equations) where radial symmetry is present.

D. Equivalence with Lizorkin-Triebel Spaces

Theorem 5.1: Under $\Delta_2$ conditions, the Orlicz-Lizorkin-Triebel space with $q=2$ coincides with the Bessel-Orlicz space:
$$F^{s,A,2}(\mathbb{R}^n) = H^{s,A}(\mathbb{R}^n)$$
Significance: This unifies the potential space approach with the Littlewood-Paley frequency decomposition approach in the Orlicz setting.

E. Atomic Decomposition

Theorems 5.3 & 5.4: The authors provide a characterization of $F^{s,A,q}$ spaces via atomic decomposition. Any function in the space can be represented as a sum of $C^m$-atoms (localized, smooth functions with vanishing moments) weighted by coefficients in an Orlicz sequence space.
Significance: Atomic decompositions are fundamental tools for proving boundedness of operators, interpolation, and duality in function spaces.

4. Significance and Impact

• Generalization: The paper successfully removes the restriction to power-law growth ($t^p$) in potential theory, allowing for the analysis of problems with logarithmic, exponential, or mixed growth behaviors.
• Methodological Innovation: The proof techniques do not rely on the homogeneity of the modulars (a key feature of $L^p$ spaces). Instead, they use the specific properties of the Bessel kernel and singular integrals adapted to Orlicz spaces, making the results robust for a wider class of Young functions.
• Applications: The results provide the theoretical foundation for studying non-linear partial differential equations (PDEs) with non-standard growth conditions, particularly those involving fractional operators and non-local interactions. The Strauss lemma specifically aids in handling critical cases in variational problems involving radial symmetry.
• Completeness: By establishing equivalences, embeddings, and atomic decompositions, the paper constructs a comprehensive theory of Orlicz potential spaces comparable to the well-developed $L^p$ theory.

In summary, Ochoa and Salort have extended the classical framework of Bessel potentials and Lizorkin-Triebel spaces to the Orlicz setting, proving that the fundamental structural relationships between these spaces hold even when the underlying integrability is governed by general convex functions.