Calderón–Zygmund Theory on the Unit Ball

• Calderón–Zygmund theory is a framework extending singular integral analysis to the unit ball via hyperbolic metrics, dyadic decompositions, and kernel estimates.
• The theory provides precise Lp, Lorentz, and weak-type endpoint estimates for operators like Bergman projections and nonlinear n-Laplace systems.
• Its analytical tools, including adapted dyadic cubes and stopping time arguments, enable refined control over mapping properties in both linear and nonlinear settings.

Calderón–Zygmund theory on the unit ball concerns the extension of singular integral analysis and weak-type endpoint estimates to geometric measure spaces such as the Euclidean unit ball $B_n \subset \mathbb{R}^n$, equipped with appropriate metrics and measure. On $B_n$, Calderón–Zygmund theory provides the analytical machinery for proving mapping properties of important operators—including Bergman projections for spaces of hyperbolic harmonic functions and nonlinear $n$-Laplace systems—yielding precise $L^p$, Lorentz, and weak-* endpoint estimates. This involves constructing adapted dyadic decompositions, obtaining singular kernel bounds with respect to the (pseudo-)hyperbolic distances, and employing sophisticated decomposition and stopping time arguments within the particular geometric setting.

1. $\mathcal{H}$-Harmonic Bergman Spaces and Projection

Let $B_n = \{x \in \mathbb{R}^n: |x| < 1\}$, with normalized Lebesgue measure $\nu(B_n) = 1$. Equip $B_n$ with the hyperbolic (Poincaré) metric

$ds^2 = \frac{4}{(1-|x|^2)^2} \sum_{i=1}^n dx_i^2,$

and associated Laplace–Beltrami operator $\Delta_h$. A function $f \in C^2(B_n)$ is $\mathcal{H}$-harmonic if $\Delta_h f \equiv 0$; define the space

$$\mathcal{H}(B_n) = \{f \in C^2(B_n) : \Delta_h f = 0\}.$$

The $\mathcal{H}$-harmonic Bergman spaces for $1 \le p < \infty$ are

$$\mathcal{B}^p = \{ f \in \mathcal{H}(B_n): \|f\|_{L^p(B_n)} < \infty \}.$$

$\mathcal{B}^2$ is a reproducing-kernel Hilbert space with kernel $\mathcal{R}(x, y)$, satisfying

$f(x) = \int_{B_n} f(y) \mathcal{R}(x, y) d\nu(y)$

for $f \in \mathcal{B}^2$ and $x \in B_n$. The corresponding orthogonal projection $P:L^2(B_n) \to \mathcal{B}^2$ is given by

$$(Pf)(x) = \int_{B_n} f(y) \mathcal{R}(x, y) d\nu(y),$$

and extends pointwise to all $f \in L^1(B_n)$ (Zhang, 5 Jan 2026).

2. Kernel Structure and Calderón–Zygmund Estimates

Define the "pseudo-distance"

$[x, y]^2 = |x - y|^2 + (1 - |x|^2)(1 - |y|^2),$

which is equivalent to the hyperbolic distance on $B_n$. A key result (Ureyen 2023) asserts: $$|\mathcal{R}(x, y)| \le \frac{C}{[x, y]^n}, \qquad |\nabla_x \mathcal{R}(x, y)| \le \frac{C}{[x, y]^{n+1}}$$ for $x \ne y$ and some $C$ depending only on $n$. Thus, $\mathcal{R}(x, y)$ is a standard Calderón–Zygmund kernel on $(B_n, \rho, d\nu)$ with $\rho(x, y) = |x - y|$ (Zhang, 5 Jan 2026).

3. Dyadic Decomposition and Stopping-Time Structure

An essential technical tool is an adapted dyadic decomposition of $(B_n, \rho, \nu)$. By results of Hytönen–Kairema, there exists a countable family of Borel sets ("dyadic cubes") $\mathscr{D} = \{ Q_{k, i} \}$, with associated points $x_{k, i} \in B_n$ and constants $0 < \eta < 1$, $0 < \kappa_0 < \kappa_1 < \infty$, such that:

• $$B(x_{k,i}, \kappa_0 \eta^k) \subset Q_{k,i} \subset B(x_{k,i}, \kappa_1 \eta^k)$$,
• For fixed $k$, the cubes $\{Q_{k,i}\}_i$ are pairwise disjoint and their union is $B_n$,
• If $k < \ell$, then each $Q_{\ell, j}$ is either contained in some $Q_{k,i}$ or disjoint from it.

These families provide the basis for the Calderón–Zygmund decomposition in this metric space, enabling local averaging and control of function oscillations in the setting of the unit ball (Zhang, 5 Jan 2026).

4. Calderón–Zygmund Decomposition and Weak-Type (1,1) Estimates

Given $f \in L^1(B_n)$ and $t > 0$, the decomposition yields a union $\Omega = \bigcup_j Q_j$ with $F = B_n \setminus \Omega$, such that on each bad cube $Q_j$,

$$\int_{Q_j} |f| d\nu > t \nu(Q_j), \qquad \int_{Q_j} |f| d\nu \le C_1 t \nu(Q_j),$$

and $|f(x)| \le t$ for almost every $x \in F$. Moreover, the measure of the "bad" set is controlled as $\nu(\Omega) \le \|f\|_{L^1}/t$. This decomposition enables the proof that the $\mathcal{H}$-harmonic Bergman projection $P$ is of weak type (1,1): $$\nu\left(\{x \in B_n : |Pf(x)| > t\}\right) \le \frac{C}{t} \|f\|_{L^1(B_n)}$$ for some $C$ independent of $f$ and $t$. The proof exploits the size and smoothness estimates of $\mathcal{R}(x, y)$ and the decomposition structure to separately bound contributions from the "good" (averaged) and "bad" (cancellation) parts (Zhang, 5 Jan 2026).

5. Application to Limiting Nonlinear Calderón–Zygmund Theory

Limiting Calderón–Zygmund theory extends to nonlinear PDEs, such as rotated $n$-Laplace systems: $$-\mathrm{div}\left( Q|\nabla u|^{n-2} \nabla u\right) = \mathrm{div}(G), \qquad u \in W^{1,n}(B_1; \mathbb{R}^N),\ Q \in W^{1,n}(B_1; SO(N)),$$ with $$G \in L^{\left( \frac{n}{n-1},q \right)} (B_1; \mathbb{R}^n \otimes \mathbb{R}^N)$$ for $0 < q < \frac{n}{n-1}$. For data $G$ in Lorentz spaces and small norm conditions on $Q$, the solution $u$ satisfies the Lorentz-scale estimate

$\nabla u \in L^{(n,\,q(n-1))}_{\mathrm{loc}}(B_1),$

together with

$$\|\nabla u\|_{L^{(n,\,q(n-1))}(B_r)} \le C\left[ \|\nabla u\|_{L^n(B_{2r})}^\alpha + \|G\|_{L^{(\frac n{n-1},\,q)}(B_{2r})} \right]$$

for suitable $0 < r < 1/2$ and $\alpha = \frac{q(n-1)}{n}$ (Martino et al., 2024). This analysis applies maximal-function/level-set decompositions, reverse Hölder inequalities, and level-set summability arguments, capitalizing on the Lorentz-space mapping properties of the Hardy–Littlewood maximal operator.

6. Endpoint and Limiting Estimates; No Further Hypotheses

For the standard $n$-Laplacian $\Delta_n u \in \mathcal{H}^1$, where $\mathcal{H}^1$ is the Hardy space, one obtains the limiting "endpoint" integrability

$\nabla u \in L^{(n, n-1)}_{\mathrm{loc}}$

with quantitative control depending on the Hardy norm of the right-hand side. In both the linear and nonlinear settings, these results extend the classical $W^{1, n} \to L^{(n, \infty)}$ endpoint theory to the Lorentz refinements and fully nonlinear regime. Notably, these theorems for the unit ball require no additional assumptions such as Muckenhoupt weights, boundary regularity, or smoothness beyond normalization of Lebesgue measure and small data for coefficients in the nonlinear case (Zhang, 5 Jan 2026, Martino et al., 2024). The central analytic inputs remain the Calderón–Zygmund kernel bounds, dyadic decomposition adapted to the metric-measure geometry, and standard decomposition/level-set iteration arguments.