Riesz–Feller Fractional Differential Operator

• Riesz–Feller fractional differential operator is a nonlocal pseudo-differential operator defined via Fourier multipliers with parameters α > 0 and skewness θ, enabling analysis of anisotropic and fractional diffusion.
• It serves as the generator of strictly α‐stable Lévy processes and drives fractional evolution equations, offering analytic semigroup generation and robust spectral representations.
• Applications span anomalous diffusion, hypercomplex analysis, and higher-order PDE frameworks, with numerical methods leveraging spectral discretizations to solve complex fractional models.

The Riesz–Feller fractional differential operator is a nonlocal pseudo-differential operator generalizing the classical Riesz operator, parameterized by a real order $\alpha > 0$ and a real skewness parameter $\theta$. It acts as the generator for strictly $\alpha$-stable (possibly skewed) Lévy processes and forms the core of fractional evolution equations, incorporating direction-dependent asymmetry. In modern analysis, it is often constructed via its Fourier multiplier, producing analytic semigroups, contour-integral fundamental solutions, and broad connections to higher-order Dirac-type operators, hypercomplex function theory, and spectral theory (Faustino, 2021).

1. Pseudo-Differential Definition and Symbol

Let $f \in \mathcal{S}(\mathbb{R}^n)$ and define the Euclidean Laplacian $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$ as well as the Riesz–Hilbert transform $\mathcal{H}$ with Fourier multiplier $h(\xi) = -i \xi / |\xi|$ (Faustino, 2021). The Riesz–Feller operator $D^{\alpha,\theta}$ is defined pseudo-differentially as: $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$ where $\mathcal{F}$ denotes the Fourier transform. The full symbol is: $\theta$0 For $\theta$1, one recovers the classical Riesz–Feller multiplier $\theta$2 (Faustino, 2021, Saxena et al., 2015, Rosu et al., 2020, Baqer et al., 2017).

2. Admissibility Conditions and Parameter Ranges

The parameters are subject to admissibility regimes ensuring analytic semigroup generation and L\'evy process correspondence:

• Order: $\theta$3 (more generally $\theta$4, $\theta$5 for higher-order theory).
• Skewness: $\theta$6 for stability and convergence (Faustino, 2021, Achleitner et al., 2014).
• For the classical range: $\theta$7 (Achleitner et al., 2014, Saxena et al., 2015). The phase $\theta$8 must remain in the closed right half-plane to guarantee convergence of the L\'evy-type kernel representations and positivity of the semigroup kernel for $\theta$9 (Faustino, 2021).

3. Integral Representations and Special Cases

For $\alpha$0, $\alpha$1, $\alpha$2, the real-space singular integral form is: $\alpha$3 (Saxena et al., 2015, Achleitner et al., 2014, Beghin, 2016). In higher dimensions, one can write $\alpha$4 as a linear combination of hypersingular Riesz-type integrals: $\alpha$5 with $\alpha$6 the classical Riesz transform kernel $\alpha$7 (Faustino, 2021).

For $\alpha$8, one recovers the symmetric fractional Laplacian $\alpha$9 (Saxena et al., 2015).

4. Operator-Theoretic Properties and Functional Analysis

• Linearity and pseudo-differential structure: $f \in \mathcal{S}(\mathbb{R}^n)$0 is linear and acts continuously on Sobolev or suitable Bochner spaces for $f \in \mathcal{S}(\mathbb{R}^n)$1.
• Ellipticity and homogeneity: $f \in \mathcal{S}(\mathbb{R}^n)$2, which implies $f \in \mathcal{S}(\mathbb{R}^n)$3-homogeneity and ensures analytic semigroup generation (Faustino, 2021).
• Markovian versus non-Markovian dynamics: For $f \in \mathcal{S}(\mathbb{R}^n)$4 and $f \in \mathcal{S}(\mathbb{R}^n)$5, the generated semigroup preserves positivity and corresponds to symmetric $f \in \mathcal{S}(\mathbb{R}^n)$6-stable processes; for $f \in \mathcal{S}(\mathbb{R}^n)$7 or $f \in \mathcal{S}(\mathbb{R}^n)$8, this may fail, resulting in skewed $f \in \mathcal{S}(\mathbb{R}^n)$9-stable Feller processes and loss of positivity (Faustino, 2021).
• Scaling: $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$0 (Achleitner et al., 2014).
• Self-adjointness: Only for $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$1; otherwise, $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$2 is non-symmetric and not self-adjoint (Achleitner et al., 2014, Saxena et al., 2011).
• Domain of definition: Schwartz space $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$3, extendable to $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$4, certain Sobolev or distributional spaces (Achleitner et al., 2014, Saxena et al., 2011, Baqer et al., 2017).

5. Fundamental Solutions and Semigroup Structure

Consider the fractional evolution equation (space-fractional Dirac-type): $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$5 (Faustino, 2021). In Fourier space, the solution evolves according to the multiplier $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$6: $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$7 which can be decomposed via idempotents $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$8: $\Delta=\sum_{j=1}^{n} \partial_{x_j}^2$9 and the inverse Fourier transform yields, for $\mathcal{H}$0,

$\mathcal{H}$1

where $\mathcal{H}$2 is the symmetric $\mathcal{H}$3-stable heat kernel (Faustino, 2021, Achleitner et al., 2014, Oraby et al., 2022).

6. Connections to Higher-Order PDEs, Hypercomplex Analysis, and Spectral Theory

The Riesz–Feller operator unifies frameworks for polyharmonic ($\mathcal{H}$4, even-order) and Dirac-type ($\mathcal{H}$5, odd-order) flows:

• Even powers: $\mathcal{H}$6 for $\mathcal{H}$7
• Odd powers: $\mathcal{H}$8 for $\mathcal{H}$9 Thus, $h(\xi) = -i \xi / |\xi|$0 provides a pseudo-differential framework for higher-order heat equations and their hypercomplex (Clifford) generalizations (Faustino, 2021).

In Paley–Wiener-type theorems, the operator defines Hardy-type spaces in upper and lower half-spaces via semigroup representations, and the support of the Fourier transform of an $h(\xi) = -i \xi / |\xi|$1 function is characterized in terms of exponential Bernstein-type bounds involving powers of $h(\xi) = -i \xi / |\xi|$2 (Bernstein et al., 2024).

7. Probabilistic Interpretation and Applications

$h(\xi) = -i \xi / |\xi|$3 is the generator of strictly $h(\xi) = -i \xi / |\xi|$4-stable Lévy processes with skewness $h(\xi) = -i \xi / |\xi|$5, with characteristic exponent $h(\xi) = -i \xi / |\xi|$6. The fundamental solution of

$h(\xi) = -i \xi / |\xi|$7

is the probability density of $h(\xi) = -i \xi / |\xi|$8, where $h(\xi) = -i \xi / |\xi|$9 is the $D^{\alpha,\theta}$0-stable process (Oraby et al., 2022). In the time-fractional case with Caputo derivative, the Green kernel becomes a convolution with the Mittag-Leffler function. Monte Carlo simulation of such PDEs proceeds by subordination methods involving sampling of stable increments and inverse subordinators (Oraby et al., 2022).

8. Numerical Methods and Spectral Discretizations

For $D^{\alpha,\theta}$1, spectral and finite-difference methods decompose the fractional Laplacian in eigenfunction bases. For $D^{\alpha,\theta}$2, direct integral-operator discretization is necessary. In one dimension, pseudospectral methods based on decomposition into Weyl–Marchaud derivatives and rational Higgins functions produce explicit formulas in terms of hypergeometric functions (Cuesta et al., 2024). Jacobi polynomial-based spectral collocation formulas for $D^{\alpha,\theta}$3 yield spectral spatial accuracy for Lévy–Feller advection-dispersion equations (Sweilam et al., 2018).

9. Fundamental Analytical and Operator Properties

Property	Description	Source
Linearity	$D^{\alpha,\theta}$4 is linear on admissible function spaces	(Achleitner et al., 2014, Saxena et al., 2011)
Scaling	$D^{\alpha,\theta}$5	(Achleitner et al., 2014)
Self-adjointness	Only for $D^{\alpha,\theta}$6; otherwise non-self-adjoint	(Achleitner et al., 2014, Saxena et al., 2011)
Domain	Schwartz, $D^{\alpha,\theta}$7, Sobolev, Bochner spaces for 1 < p < ∞	(Achleitner et al., 2014, Faustino, 2021)
Pseudo-differential form	Fourier multiplier with symbol $D^{\alpha,\theta}$8	(Faustino, 2021)
Semigroup generation	Analytic semigroup for admissible $D^{\alpha,\theta}$9	(Faustino, 2021)
Probabilistic generator	Infinitesimal generator of strictly $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$0-stable Lévy process	(Oraby et al., 2022)

10. Special Cases and Limiting Behavior

• For $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$1 and $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$2, $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$3 becomes the Laplacian.
• For $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$4, $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$5, the operator reduces (up to constant) to the first derivative.
• For $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$6, the operator is the symmetric Riesz derivative (fractional Laplacian) (Saxena et al., 2015, Saxena et al., 2011).

11. Closed-Form Solutions and Representation

The operator $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$7 allows for closed-form solutions of fractional reaction-diffusion equations and Schrödinger-type equations via Fourier-Laplace transform inversion, producing solutions in terms of Mittag-Leffler functions, Wright functions, Mellin-Barnes integrals, and Fox $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$8-functions, with direct links to the anisotropy and skewness imparted by $$D^{\alpha,\theta}f(x) = \mathcal{F}^{-1}\left[ -|\xi|^{\alpha} \exp\left( i\pi\theta h(\xi)/2 \right) \mathcal{F}f(\xi) \right](x),$$9 (Saxena et al., 2011, Saxena et al., 2014, Rosu et al., 2020, Baqer et al., 2017).

---

The Riesz–Feller fractional differential operator $\mathcal{F}$0 underpins modern nonlocal analysis, anomalous diffusion modeling, hypercomplex PDE frameworks, generalized harmonic analysis, and fractional quantum mechanics. Its formalism provides a coherent bridge between stochastic processes, functional analysis, operator theory, and numerical modeling (Faustino, 2021, Bernstein et al., 2024, Oraby et al., 2022, Cuesta et al., 2024, Sweilam et al., 2018, Saxena et al., 2014, Saxena et al., 2011).