Laplace-Transformed Backward Master Equation

• Laplace-Transformed Backward Master Equation is a formulation that uses Laplace transforms to convert time-fractional, nonlocal PDEs into parameter-dependent spatial PDEs.
• It simplifies complex stochastic models of anomalous diffusion by replacing nonlocal time derivatives with manageable spatial terms, enhancing numerical stability.
• The approach underpins local discontinuous Galerkin methods, ensuring optimal convergence rates and improved error control in simulations of fractional dynamics.

The Laplace-transformed backward master equation describes the time evolution of functionals of anomalous diffusion processes, formulated via Laplace transform techniques applied to the backward Feynman-Kac equation. Central to the analysis of functionals of stochastic dynamics under spatially varying potentials, this equation serves as the foundation for modern numerical discretizations—especially for anomalous diffusion where non-integer (fractional) time derivatives and substantial derivatives predominate. The Laplace-transformed form casts the original time-dependent, nonlocal-in-time PDE into a family of parameter-dependent spatial PDEs, substantially simplifying numerical treatment and facilitating theoretical stability and convergence analysis (Liu et al., 2022).

1. Foundational Equation and Fractional Substantial Calculus

The governing equation for the moment-generating function $G(x,t)$ of time-functionals associated with anomalous diffusion is

$$\partial_t^{\omega,V(x)} G(x,t) = \mathcal{L}\,G(x,t) - V(x)\,I_t^{\omega,V(x)} G(x,t),$$

where $x\in\Omega\subset\mathbb{R}^d$, $0 < t \leq T$, $\mathcal{L}$ denotes the generator of the underlying spatial process (e.g., for Brownian motion, $\mathcal{L} = \Delta$), $V(x)$ is a spatially varying potential, $I_t^{\omega,V}$ is a generalized fractional integral, and $\partial_t^{\omega,V}$ is the corresponding generalized (substantial) derivative. Initial data $G(x,0) = G_0(x)$ and appropriate boundary conditions (periodic or Dirichlet in $x$) are prescribed.

Specializing to Mittag-Leffler waiting distributions ($\omega(\tau) = \tau^{-\alpha}/\Gamma(1-\alpha)$ with $0<\alpha<1$), the equation admits an equivalent Caputo-type fractional substantial derivative form:

${}_0^C D_t^{\alpha,V(x)} G(x,t) = \Delta G(x,t),$

where the substantial Caputo derivative is given by

$${}_0^C D_t^{\alpha,V}u(t) = e^{-V t}\,{}_0^C D_t^\alpha\{e^{V t} u(t)\},$$

and ${}_0^C D_t^\alpha$ denotes the classic Caputo derivative.

2. Definition and Properties of the Fractional Substantial Derivative

Two fractional substantial derivatives govern the analysis:

• Riemann–Liouville version:

$${}_0^R D_t^{\alpha,V}u(t) = e^{-V t}\,{}_0^R D_t^\alpha\{e^{V t} u(t)\},$$

where $${}_0^R D_t^\alpha u(t) = \frac{1}{\Gamma(1-\alpha)}\frac{d}{dt}\int_0^t (t-s)^{-\alpha} u(s) ds$$.

• Caputo version:

${}_0^C D_t^{\alpha,V}u(t) = e^{-V t}\,\frac{1}{\Gamma(1-\alpha)}\int_0^t (t-s)^{-\alpha} \frac{d}{ds}\left[e^{V s} u(s)\right] ds = {}_0^I_t^{1-\alpha,V}[u'(t)].$

A key Laplace transform property (Lemma 2.1 in (Liu et al., 2022)) is:

$$\mathcal{L}\{{}_0^C D_t^{\alpha, V} u(t)\}(s) = (s + V)^\alpha U(s) - (s+V)^{\alpha-1}u(0)$$

for the Caputo variant with $u(0)=u_0$, where $U(s) = \mathcal{L}\{u\}(s)$. For the associated fractional substantial integral:

$$\mathcal{L}\{I_t^{\alpha,V}u(t)\}(s) = (s+V)^{-\alpha}U(s).$$

3. Laplace Transformation of the Backward Equation

Applying the Laplace transform in time to the substantial Caputo equation yields:

$$(s+V(x))^{\alpha}\,\hat{G}(x,s) - (s+V(x))^{\alpha-1} G(x,0) = \Delta \hat{G}(x,s),$$

where $\hat{G}(x,s) = \mathcal{L}\{G(x,t)\}(s)$. For a general generator $\mathcal{L}$, replace $\Delta \to \mathcal{L}$. The result is a family of parameter-dependent elliptic PDEs indexed by the Laplace variable $s$.

Rearranging, the standard Helmholtz-type form emerges:

$$- \Delta \hat{G}(x,s) + (s + V(x))^{\alpha} \hat{G}(x,s) = - (s + V(x))^{\alpha - 1} G_0(x),$$

where the right-hand side source is determined by the initial data $G(x,0) = G_0(x)$. Transformed boundary conditions (e.g., periodicity in $x$) are imposed correspondingly.

4. Role in Local Discontinuous Galerkin Discretization

The Laplace-transformed backward master equation is recast as a stationary, parameter-dependent elliptic system, forming the basis of local discontinuous Galerkin (LDG) methods for spatial discretization. The procedure entails:

• First-order system conversion: Set $u(x) = \hat{G}(x,s)$, $p(x) = \nabla u(x)$. The system becomes: \begin{align*} -\nabla\cdot p + (s+V)^{\alpha} u &= -(s+V)^{\alpha-1} G_0, \ p - \nabla u &= 0. \end{align*}
• Finite element variational statement: On each grid element $\Omega_{ij}$, seek $(u_h, p_h)$ in $[Q^k]^{d+1}$ so that:

$$\int_{\Omega_{ij}} \left[-p_h \cdot \nabla v + (s+V)^{\alpha} u_h v\right] + \int_{\Omega_{ij}} p_h \cdot w - \int_{\Omega_{ij}} u_h \nabla \cdot w + \text{boundary-flux terms} = \int_{\Omega_{ij}} \left[-(s+V)^{\alpha-1} G_0\right] v$$

for test functions $v$, $w$.

• Numerical fluxes: On each mesh face, a "generalized alternating" flux with weights $\sigma_1\neq1/2$, $\sigma_2\neq1/2$ is deployed: \begin{align*} \hat{u} &= \sigma_1\,u_h^- + (1 - \sigma_1) u_h^+ \ \hat{p} n_x &= (1-\sigma_1) p_h^- n_x + \sigma_1 p_h^+ n_x \end{align*} Analogous definitions hold for the $y$-direction using $\sigma_2$.
• Stability: The choice of alternating flux ensures the interior bilinear form vanishes ($B(u_h,u_h;p_h,p_h)=0$ by Lemma 3.1), and, since $(s+V)^{\alpha}$ is positive, yields $L^2$-stability for the discrete solution $u_h$.

The discrete framework thereby leverages the Laplace-transformed form, yielding stability and optimal convergence rates $O(h^{k+1})$ in space (Liu et al., 2022).

5. Implications for Numerical Analysis and Application Domains

The Laplace-transformed formulation transforms the time-fractional, nonlocal problem into an elliptic one, facilitating the design of LDG schemes with provable stability and optimal convergence. Key consequences include:

• Error control near initial singularities: The $L^1$ time-stepping scheme on graded meshes is utilized to manage solution singularities proximate to $t=0$.
• Fully discrete scheme performance: The theoretical results for the semi-discrete (in space) system extend to the full discretization (space and time), yielding optimal rates:

$$O\left(h^{k+1} + \tau^{\min\{2-\alpha, \gamma\delta\}}\right).$$

• Boundary and flux choices impact: The choice of central versus alternating fluxes affects both convergence order and matrix conditioning.

An immediate application domain is the computation of functionals for anomalous diffusion, including Eulerian descriptions of subdiffusive transport in complex media, as detailed in the reference work (Liu et al., 2022). This framework provides a systematic pathway for addressing a broad class of backwards-time, nonlocal, stochastic PDEs of physical and mathematical relevance.

6. Context and Significance Within Anomalous Diffusion Theory

The Laplace-transformed backward master equation is essential in the study of anomalous diffusion, as it mathematically formalizes the evolution of time-integrated functionals subjected to fractional kinetics and spatial heterogeneity. The associated substantial fractional calculus rigorously captures memory effects and field-induced modulation, which are pivotal in subdiffusion and continuous time random walk models. By enabling tractable and accurate numerical approaches, notably through the LDG methodology, the Laplace-transformed equation has advanced both the theoretical and computational study of nonlinear transport phenomena and stochastic functional analysis in complex systems (Liu et al., 2022).