Spectral Fractional Operators

• Spectral fractional operators are mathematical constructs that generalize classical differential operators using non-integer powers via spectral methods.
• They provide key findings such as generalized Weyl laws for eigenvalue asymptotics and efficient spectral discretization techniques for numerical analysis.
• Extensions include nonlinear, nonlocal, and superposition operators, offering robust tools for advanced analysis in PDEs, quantum mechanics, and optimization.

Spectral fractional operators are mathematical constructs that generalize classical differential and pseudo-differential operators by allowing their (possibly non-integer) powers and more general functional calculus, with definitions rooted in spectral theory. These operators encompass the fractional Laplacian, fractional derivatives, nonlocal elliptic operators, and various nonlinear and superposed structures, playing fundamental roles in analysis, PDE theory, quantum mechanics, optimization, and mathematical physics.

1. Definitions and Classes of Spectral Fractional Operators

A spectral fractional operator arises by applying a functional calculus—often fractional powers or more general functions—to a self-adjoint or sectorial operator $L$ on a Hilbert space, typically the Laplacian or more general elliptic operator, via its spectral decomposition. In a bounded domain $\Omega\subset\mathbb{R}^d$ with Dirichlet conditions, one sets

$$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$

where $-\Delta\varphi_j=\lambda_j\varphi_j$, $\{\varphi_j\}$ forms an orthonormal basis for $L^2(\Omega)$ and $\{\lambda_j\}$ are the eigenvalues. For more general nonlocal or pseudo-differential operators, a similar construction applies: let $T(\xi)$ be the symbol, then the operator $L_T$ is defined on $L^2(\Omega)$ by

$\Omega\subset\mathbb{R}^d$0

with form domain determined by the integrability against the weight $\Omega\subset\mathbb{R}^d$1 (Geisinger, 2013).

Classes of spectral fractional operators include:

• Fractional Laplacians: $\Omega\subset\mathbb{R}^d$2, leading to $\Omega\subset\mathbb{R}^d$3 for $\Omega\subset\mathbb{R}^d$4.
• Pseudo-differential operators of fractional and mixed order: with non-homogeneous symbols $\Omega\subset\mathbb{R}^d$5 such as $\Omega\subset\mathbb{R}^d$6, $\Omega\subset\mathbb{R}^d$7.
• Weyl, Riemann-Liouville, Caputo, and other nonlocal fractional derivatives and integrals in continuous or discrete settings.
• Nonlinear and superposition operators integrating operators of different (fractional) orders against a measure, including sums such as $\Omega\subset\mathbb{R}^d$8 and even infinite series or "superpositions" parameterized by a Borel measure (Aikyn et al., 4 Nov 2025, Dipierro et al., 15 Apr 2025).
• Weighted, directional, or non-symmetric fractional operators (e.g., involving directional integrals, weighted Weyl calculus, or compositions of left and right-sided derivatives).

2. Spectral Asymptotics and Weyl's Law

For self-adjoint spectral fractional operators with suitable homogeneity, the counting function $\Omega\subset\mathbb{R}^d$9 for eigenvalues below $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$0 obeys a generalized Weyl law. For

$$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$1

$$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$2

as $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$3, where $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$4 and $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$5 is the leading homogeneous symbol (Geisinger, 2013). Key special cases:

• Dirichlet Laplacian: $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$6, $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$7, so $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$8.
• Fractional Laplacian: $$(-\Delta)^s u = \sum_{j=1}^\infty \lambda_j^s \langle u, \varphi_j \rangle \varphi_j, \quad s>0$$9, $-\Delta\varphi_j=\lambda_j\varphi_j$0, so $-\Delta\varphi_j=\lambda_j\varphi_j$1 with $-\Delta\varphi_j=\lambda_j\varphi_j$2 computed explicitly via Gamma functions.

For nonhomogeneous symbols or superpositions, the highest-order term governs the asymptotic, and the spectral asymptotics reduce to those of the Laplacian or the dominating fractional power (Geisinger, 2013, Dipierro et al., 15 Apr 2025). In nonsymmetric or more exotic nonlocal settings, asymptotic behavior is often less explicit, but similar power-law growth is expected and numerically observed (Deng et al., 2022).

3. Operator Structure: Domains, Regularity, and Spectral Projections

The domain of a spectral fractional operator is determined by the associated quadratic form or the relevant spectral characterization. For $-\Delta\varphi_j=\lambda_j\varphi_j$3, the domain is the spectral Sobolev space

$-\Delta\varphi_j=\lambda_j\varphi_j$4

with norm

$-\Delta\varphi_j=\lambda_j\varphi_j$5

which for $-\Delta\varphi_j=\lambda_j\varphi_j$6 coincides with $-\Delta\varphi_j=\lambda_j\varphi_j$7 (Otarola et al., 11 Feb 2026). General pseudo-differential or superposition operators may require more subtle function spaces, especially in the presence of singular measure-valued data.

Eigenfunctions for spectral fractional operators are smooth in the interior, but possess characteristic boundary singularities dictated by the transmission property and principal symbol. For $-\Delta\varphi_j=\lambda_j\varphi_j$8 an eigenfunction of $-\Delta\varphi_j=\lambda_j\varphi_j$9 with Dirichlet data on a smooth domain,

$\{\varphi_j\}$0

where $\{\varphi_j\}$1 is distance to the boundary (Grubb, 2014).

Spectral projections and functional calculus for these operators are implemented via the eigenbasis, enabling definitions of general $\{\varphi_j\}$2 for Borel/Holomorphic $\{\varphi_j\}$3 and corresponding spectral mapping theorems for the operator spectrum (Dorrego, 5 Jan 2026).

4. Extensions: Superposition, Nonlocality, and Nonsymmetry

Superposition operators form a prominent modern extension: given a signed Borel measure $\{\varphi_j\}$4 on $\{\varphi_j\}$5, the operator

$\{\varphi_j\}$6

acts as a continuous superposition of fractional Laplacians. The spectrum is discrete provided $\{\varphi_j\}$7 carries mass in a high enough order and $\{\varphi_j\}$8 is not too large; the principal frequency satisfies a Faber-Krahn inequality, is isolated, and eigenfunctions are regular and bounded. Negative parts in $\{\varphi_j\}$9 can break maximum principles and must be dominated by the positive part on high orders for spectral stability (Dipierro et al., 15 Apr 2025, Aikyn et al., 4 Nov 2025). Variational characterizations and Courant–Fischer min-max formulas remain valid in the presence of suitable positivity and symmetry.

Nonlocal and nonsymmetric fractional elliptic operators (e.g., compositions of left/right Riemann-Liouville derivatives) present spectral problems with complex-valued spectra, limited real eigenvalues, and novel challenges regarding completeness of eigenfunctions and explicit spectral characterization; the set of eigenvalues can be described analytically for special cases, but the general problem is often open (Deng et al., 2022).

Weighted and time-dependent fractional operators (e.g., the weighted Weyl calculus for aging/non-Markovian materials) admit a continuous spectrum, explicit spectral measure, and functional calculus via weighted Fourier transforms; weighted Mittag-Leffler functions form their generalized eigenfunctions, encapsulating phenomena like the "amnesia effect" in rapidly aging viscoelastic media (Dorrego, 5 Jan 2026).

5. Discretizations, Numerical Realizations, and Spectral Methods

Spectral fractional operators admit discretizations amenable to high-accuracy computation:

• Spectral methods: Use eigenbasis expansions or mapped Chebyshev/Jacobi/Laguerre polynomials adapted to the nonlocal/fractional structure; spectral collocation and Galerkin approaches diagonalize the operator, facilitating efficient solution of PDEs and eigenvalue problems (Sheng et al., 2020, Antil et al., 2017, Liu et al., 24 Jun 2025, Fatone et al., 2014).
• Finite element schemes: Discrete Laplacians with computed eigenpairs are used to construct fractional powers in finite-dimensional spaces, with convergence rates governed by solution regularity and mesh size; Balakrishnan integral representations and quadrature reduce fractional inversion to finite system solves (Otarola et al., 11 Feb 2026).
• Discrete fractional Sturm-Liouville (DFSL): Admitting self-adjoint, real-spectral structure analogous to their continuous fractional and classical integer-order counterparts (Bas et al., 2017).

In all cases, spectral fractional discretizations maintain rapid, often exponential, convergence for analytic data with accurate matrix representations, well-conditioned numerics, and effective algorithms for both linear and nonlinear, local and nonlocal, time-dependent and optimization-driven applications.

6. Nonlinear, Geometric, and Algebraic Extensions

Nonlinear spectral fractional operators, such as superpositions of nonlinear fractional $L^2(\Omega)$0-Laplacians, possess variationally characterized spectra, isolated positive principal eigenvalues, bounded and regular eigenfunctions, and satisfy analogs of mountain-pass theorems and geometric inequalities, such as the Faber–Krahn inequality (Aikyn et al., 4 Nov 2025). In geometric and conformal settings, spectral definitions of fractional GJMS operators on Poincaré–Einstein manifolds yield conformally covariant boundary operators and sharp trace inequalities, with spectra controlled via extension problems (Case, 2015).

On the algebraic and integrable-systems side, algebras of commuting spectral fractional differential operators are classified via extensions of the Burchnall–Chaundy theorem, Sato's Grassmannian, and Krichever theory, with maximal commutative subalgebras arising from rational data in jet-bundle constructions and spectral fields (Casper et al., 2021).

7. Summary Table of Spectral Fractional Operator Types

Operator Class	Principal Symbol / Definition	Core Spectral Feature
Fractional Laplacian	$L^2(\Omega)$1	Weyl law: $L^2(\Omega)$2
Pseudo-differential (homogeneous)	$L^2(\Omega)$3, $L^2(\Omega)$4	Weyl law: leading $L^2(\Omega)$5 term governs asymptotics
Superposition / Mixed Order	$L^2(\Omega)$6	Discrete spectrum; highest-order governs leading term
Nonlocal nonsymmetric (e.g., RL)	$L^2(\Omega)$7	Cone of complex eigenvalues; positive principal; open asymptotics
Weighted Weyl / Memory-Aging	$L^2(\Omega)$8	Continuous spectrum $L^2(\Omega)$9; eigenfunctions $\{\lambda_j\}$0
Discrete Fractional Sturm-Liouville	Fractional finite-difference	Real, discrete spectrum; self-adjointness, completeness
Nonlinear fractional $\{\lambda_j\}$1-Laplacian	$\{\lambda_j\}$2	Variational eigenvalues, isolation, maximum principles

Spectral fractional operators thus constitute a broad, flexible class of linear and nonlinear, local and nonlocal operators whose spectral properties underlie much of contemporary analytic, geometric, and numerical PDE theory, as well as numerous applications in the mathematical sciences (Geisinger, 2013, Grubb, 2014, Antil et al., 2017, Otarola et al., 11 Feb 2026, Aikyn et al., 4 Nov 2025, Dipierro et al., 15 Apr 2025, Deng et al., 2022, Dorrego, 5 Jan 2026, Casper et al., 2021).