Multilinear Calderón-Zygmund Theory

• Multilinear Calderón-Zygmund Theory extends classical singular integrals to operators acting on tuples of functions, using spectral multipliers and harmonic analysis on Lie groups.
• It establishes sharp regularity thresholds, endpoint bounds, and weighted inequalities through techniques like sparse domination and functional calculus.
• Applications range from dispersive PDEs and Bochner-Riesz means to noncommutative analysis, with ongoing research addressing minimal smoothness in complex group settings.

Multilinear Calderón-Zygmund Theory is an extension of classical singular integral theory to operators acting on tuples of functions, rooted in harmonic analysis on stratified Lie groups, and driven by questions of minimal regularity for boundedness properties of spectral multipliers. In the multilinear framework, Calderón-Zygmund-type reasoning provides a foundation for endpoint and weighted bounds for classes of operators, including oscillating and rough spectral multipliers, on a broad range of non-commutative and inhomogeneous group settings. Rigorous developments in the last decade have established sharp differentiability thresholds, sparse domination principles, and weighted extrapolation techniques for multilinear operators on both polynomial-growth and exponential-growth groups.

1. Structural Foundations and Setting

The classical Calderón-Zygmund theory hinges on kernels $K(x,y)$ satisfying singularity, cancellation, and smoothness conditions—controlling the action of convolution-type operators on $L^p$ spaces via singular integrals. In the multilinear case, this extends to operators of the form

$$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$

with $K$ singular at the diagonal $y_1 = y_2 = \dots = y_m = x$. On stratified Lie groups and their extensions, such as 2-step nilpotent, Heisenberg, and Métivier groups, the relevant operators are obtained as spectral multipliers of sub-Laplacians or Rockland operators, via the functional calculus:

$F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$

with $L$ a sub-Laplacian and $E_L$ the spectral measure. This structure allows the translation of multiplier smoothness into size and regularity estimates for the associated convolution kernel $K_{F(L)}$.

Multilinear theory also interfaces with sparse domination principles: a multilinear operator $T$ is controlled by a sparse bilinear form, yielding sharp $L^p$0-weighted bounds under appropriate moment and cancellation conditions on the kernel (Ghosh et al., 2023).

2. Sharp Thresholds for Spectral Multiplier Regularity

A central problem is determining for which regularity $L^p$1 a function $L^p$2, satisfying a scale-invariant Sobolev-type norm

$L^p$3

ensures boundedness of $L^p$4 on $L^p$5 for all $L^p$6. For general graded groups of homogeneous dimension $L^p$7, the classical threshold is $L^p$8 (Duong et al., 2010, Cardona et al., 2016). However, major advances have established that on specific 2-step stratified groups, particularly with step-2 center dimension $L^p$9 or topological dimension $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$0, the optimal threshold drops to the Euclidean-type $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$1, with $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$2 (Martini et al., 2013, Martini, 2012).

On groups of Heisenberg type—where the commutator map $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$3 in the first stratum is diagonalizable—the sharp threshold $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$4 for Mihlin-Hörmander multipliers is strictly less than $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$5 and is best possible (Bramati et al., 2020). Recent results further extend this threshold to certain solvable semidirect product extensions, yielding for left-invariant sub-Laplacians on $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$6 with $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$7 a direct product of Heisenberg-type and abelian groups, a threshold $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$8 for $$T(f_1, \dots, f_m)(x) = \int \dots \int K(x; y_1, \dots, y_m) \prod_{j=1}^m f_j(y_j)\, dy_1 \dots dy_m,$$9 (Martini et al., 2023).

The following table summarizes regularity thresholds for weak-type $K$0 and $K$1-boundedness of spectral multipliers $K$2 in various group classes.

Group Type	Dimension Parameter	Threshold $K$3	Reference
General stratified group	$K$4 (homogeneous)	$K$5	(Duong et al., 2010)
2-step, $K$6 or $K$7	$K$8 (topological)	$K$9	(Martini et al., 2013)
Heisenberg-type	$y_1 = y_2 = \dots = y_m = x$0 (topological)	$y_1 = y_2 = \dots = y_m = x$1	(Bramati et al., 2020)
$y_1 = y_2 = \dots = y_m = x$2, $y_1 = y_2 = \dots = y_m = x$3 Heisenberg-type	$y_1 = y_2 = \dots = y_m = x$4	$y_1 = y_2 = \dots = y_m = x$5	(Martini et al., 2023)

Reducing $y_1 = y_2 = \dots = y_m = x$6 allows for rougher multipliers, crucial for applications to PDEs and dispersive equations.

3. Multilinear Calderón-Zygmund Methods and Weighted Theory

Multilinear Calderón-Zygmund tools remain essential for handling endpoint bounds, kernel $y_1 = y_2 = \dots = y_m = x$7 estimates, and weighted inequalities. The multilinear setting requires careful control of the oscillatory and smoothing behavior of the kernel, often decomposed dyadically in frequency and spatial variables.

For operators of the form $y_1 = y_2 = \dots = y_m = x$8 with oscillatory $y_1 = y_2 = \dots = y_m = x$9, sparse domination results establish that $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$0 admits sharp (quasi-)local $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$1 bounds as soon as $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$2 satisfies suitable size and smoothness decay conditions (Ghosh et al., 2023). This enables the derivation of quantitative bounds for Muckenhoupt $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$3-weighted norms and endpoint estimates of weak-type $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$4.

4. Endpoint and Oscillating Multiplier Theorems

Oscillating spectral multipliers—relevant for dispersive PDE and time-dependent evolutions—exhibit endpoint $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$5 behavior sharply governed by the group’s structure. On Heisenberg-type groups, the spectral multiplier

$F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$6

yields strong-type $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$7 bounds whenever $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$8 provided the cutoff $F(L) = \int_0^\infty F(\lambda)\, dE_L(\lambda),$9 has Sobolev regularity $L$0, with weak-type $L$1 precisely at the endpoint (Bramati et al., 2020). The range and sharpness of these results are tightly connected to dispersive and wave equation estimates.

In spaces of exponential volume growth and nondoubling measures, as in some solvable extensions, the proof of Euclidean-type multiplier theorems necessitates lifting Plancherel and kernel estimates to geometrically-adapted hypergroup models (Martini et al., 2023), and the multilinear method relies on stability under group extensions and hypergroup convolution structures.

5. Technical Machinery: Functional Calculus and Harmonic Analysis

The analytic core of multilinear Calderón-Zygmund theory in this context is built on the explicit functional calculus for sub-Laplacians and Rockland operators:

$L$2

where $L$3 is controlled uniformly by a Sobolev norm

$L$4

For 2-step groups, joint spectral calculus with central variables leads to formulae involving Laguerre polynomials and mixed Fourier-Laguerre expansions (Martini et al., 2013, Martini, 2012). The analysis of the kernel $L$5 hinges on weighted $L$6-norm estimates and Fourier-analytic decomposition in the central frequency variable, including splitting into caps (angular sectors) and dyadic shells in the group dual, with region-dependent estimates to capture algebraic singularities and degeneracies.

Proof techniques draw from

• weighted Plancherel identities,
• $L$7 and $L$8 spatial and frequency-weighted kernel bounds,
• Calderón-Zygmund decomposition on non-doubling spaces,
• real interpolation and complex interpolation between Hardy, BMO, and $L$9 spaces,
• spectral cluster and restriction-type (Stein–Tomas) estimates for spectral projectors (Niedorf, 27 Jan 2025).

6. Applications, Examples, and Open Problems

Key Models and Examples

• Heisenberg groups: The improvement to $E_L$0 is fully sharp. Oscillating multipliers and dispersive estimates attain this bound (Bramati et al., 2020).
• Free 2-step nilpotent group $E_L$1: Recent work establishes the Euclidean-type threshold under suitable algebraic constraints (Niedorf, 27 Jan 2025).
• Heisenberg–Reiter groups: These groups serve as models for degenerate or block-diagonal structures; the block-decomposition of the commutator leads to sharp $E_L$2 thresholds for spectral multipliers (Martini, 2012, Niedorf, 27 Jan 2025).
• Seimdirect solvable extensions $E_L$3: Euclidean-type theorems with threshold $E_L$4 are achieved through geometric lifting and hypergroup analysis (Martini et al., 2023).

Applications

• Weighted $E_L$5 theory: Sharp Muckenhoupt $E_L$6 bounds are available for a variety of multipliers under Hörmander–Sobolev conditions (Duong et al., 2010).
• Oscillatory and dispersive PDE: Sharp endpoint Strichartz and solution space estimates follow from the multilinear sparse bounds and oscillatory multiplier bounds (Ghosh et al., 2023).
• Bochner-Riesz and Riesz means: Spectral multipliers for $E_L$7 and similar operators satisfy optimal boundedness thresholds, resolving conjectures for key classes of 2-step groups (Niedorf, 27 Jan 2025).

Open Problems and Future Directions

• Determination of the minimal smoothness threshold $E_L$8 for general graded or higher-step groups remains largely unresolved, with current knowledge that $E_L$9 but equality only verified in special cases (Bramati et al., 2020).
• For degenerate (non-Métivier) stratified groups and $K_{F(L)}$0 outside restricted ranges, endpoint weak-type and weighted inequalities are still open (Niedorf, 27 Jan 2025).
• Construction of multilinear Calderón-Zygmund theory adapted to non-Rockland and more singular operators, especially beyond the stratified or homogeneous class, remains an active area.
• Extending these theories to the setting of joint functional calculi beyond polynomial growth or to totally disconnected or nonunimodular settings presents analytic and representation-theoretic challenges.

7. Connections to Broader Harmonic Analysis and Operator Theory

The multilinear Calderón-Zygmund paradigm on Lie groups is connected to several major themes:

• Joint spectral multipliers and operator algebras: Functional calculi for systems of commuting self-adjoint operators are essential for analyzing the structure of noncommutative $K_{F(L)}$1-spaces and the spectral theory of weighted subcoercive systems (Martini, 2010, Martini, 2010).
• Besov and Triebel–Lizorkin scales on groups: The interplay with Littlewood–Paley, Nikolskii, and embedding theorems supports the extension of multiplier results to nonstandard function spaces (Cardona et al., 2016).
• Microlocal and geometric analysis: The kernel bounds and oscillation conditions are tightly related to the geometry of underlying group structures (e.g., center dimension, commutator degeneration, presence of flat or non-flat coadjoint orbits) (Chatzakou, 2021).
• Abstract noncommutative analysis: The operator-valued nature of group Fourier transforms and spectral projectors on von Neumann algebras leads to optimal $K_{F(L)}$2 bounds and heat kernel asymptotics (Rottensteiner et al., 2022, Akylzhanov et al., 2015).
• Extension to exponential volume growth groups: For semidirect solvable extensions, new analytic and kernel lifting methods must be developed to compensate for the loss of doubling or polynomial growth (Martini et al., 2023, Ottazzi et al., 2013).

The multilinear Calderón-Zygmund theory thus occupies a central position in modern harmonic analysis, mediating between algebraic group structures, analytic kernel estimates, and endpoint regularity phenomena for operator families across a wide range of geometric and analytic settings.