Personal Portfolio

What Are Fractional Differential Equations

Classification of Fractional Differential Equations

For fractional differential equations of the form

$Pu = f$

the type of operator $P$ (time-fractional, space-fractional, space-time coupled, or nonlocal integral), the function space for the field $u$, and the regularity of the source $f$ together determine the core properties of the equation. Time-fractional differential equations characterize long-range (memory) effects of systems through power-law memory kernels, so that the historical dependence of solutions persists with algebraic rather than exponential decay, enabling accurate descriptions of subdiffusion and ultraslow diffusion processes. Space-fractional differential equations employ heavy-tailed Levy stable distributions to model long-distance particle jumps, thereby providing the capability to capture anisotropic diffusion effects, making them particularly suitable for characterizing directional migration and anomalous transport phenomena in heterogeneous media. Under different combinations, the equation also exhibits significant differences in the degree of solution singularity at boundaries and the order of regularity gains.

Specific work includes:

1. For time-fractional operators (Caputo, Riemann-Liouville, distributed-order), establishing existence and uniqueness, maximal regularity, and $t=0$ singularity analysis for initial-boundary value problems in Sobolev and Holder spaces;
2. For space-fractional operators (integral fractional Laplacian, spectral fractional Laplacian, Riesz derivative), studying boundary $d^s$ regularity, interior Schauder estimates, and strong solution theory in fractional Sobolev spaces $H^s$ and Holder spaces $C^{2s+\alpha}$;
3. For strongly elliptic pseudodifferential operators $r^+ OP(p(x,\xi))$, deriving direct-sum decompositions of solution spaces and weighted gradient estimates in Bessel potential spaces $H_q^{a(2a+s)}$ and Holder-Zygmund spaces $C_*^{a(2a+s)}$;
4. For space-time coupled operators and nonlocal integral operators, establishing fundamental solution representations and Krylov-Safonov-type estimates in corresponding energy spaces.

Symbol Definitions

Fractional differential equations precisely model phenomena with long-range correlations or memory effects (such as viscoelastic relaxation, subdiffusion processes, and viral immune history responses) through the power-law memory kernels of time-fractional derivatives; and characterize spatially anisotropic diffusion and nonlocal interactions through the heavy-tailed Levy jumps of space-fractional derivatives. For specific modeling applications of fractional differential equations in practical engineering and scientific problems (including numerical toolkits, source code repositories, and representative papers), see Applications of Fractional Differential Equations.

Why Fractional Differential Equations Have Memory

Currently recommended research directions include: (1) Fast algorithms and reduced-order numerical methods, developing fast convolution, low-rank approximation, fast multipole, and tensor-type algorithms to address the $O(N^2)$ computational complexity arising from the nonlocality of fractional operators; (2) Inverse problems and parameter identification, simultaneously recovering fractional exponents, diffusion coefficients, and source terms from observational data, overcoming the ill-posedness caused by nonlinearity and nonlocality; (3) Physics-informed neural networks (PINN), embedding fractional differential constraints into neural network loss functions to achieve data-driven solutions and surrogate model construction; (4) High-dimensional and complex geometry problems, overcoming the curse of dimensionality and developing spectral methods, Monte Carlo methods, and deep learning approaches for high-dimensional fractional equations; (5) Variable-order and distributed-order equations, allowing the order to vary with spatial position or temporal evolution to capture heterogeneous media and multi-scale memory effects; (6) Stochastic fractional equations and uncertainty quantification, studying the regularity, numerical approximation, and Bayesian inference of fractional stochastic partial differential equations; (7) Emerging interdisciplinary applications, such as graph neural networks, medical image segmentation, spatiotemporal epidemic propagation modeling, and anomaly detection in high-frequency financial data.

Researchers

Mathematicians who made foundational contributions to the development of fractional calculus.