Original Energy-Dissipative Schemes

• Original-energy-dissipative schemes are numerical methods that preserve the physical energy decay law from dissipative PDEs at every discrete time step.
• They use advanced formulations like ETDRK, DVD, Lagrange multiplier, and iSAV approaches to ensure high-order accuracy and unconditional or near-unconditional stability.
• These schemes have practical applications in phase field, chemotaxis, flow, and electrokinetic models, providing reliable energy dissipation and structural preservation in simulations.

Original-energy-dissipative schemes are numerical time discretizations for dissipative PDEs and gradient flows that ensure exact non-increase of the original physical (Lyapunov or free) energy at the discrete level, matching the continuous dissipation law of the PDE, as opposed to merely dissipating a modified or auxiliary energy. These schemes have become central in the simulation of phase field, chemotaxis, reaction-diffusion, incompressible flow, aggregation-diffusion, and electrokinetic systems, and are now available for a broad spectrum of nonlinear, high-order, and coupled PDE models.

1. Theoretical Foundations and Definitions

A scalar gradient-flow system has the form

$u_t = -\mathcal{G} \frac{\delta E}{\delta u},$

with dissipation operator $\mathcal{G}\le 0$ and original energy functional $E$. The fundamental property is the physical energy decay law: $$\frac{d}{dt} E[u(t)] = \left( \frac{\delta E}{\delta u}, u_t \right) = - \|u_t\|^2 \le 0.$$ An original-energy-dissipative numerical scheme is a fully discrete time-stepping algorithm such that

$E[u^{n+1}] \leq E[u^n]$

for the original $E$ at every time step, possibly under a mild constraint on the step size. Importantly, this is distinct from stability proofs that guarantee decay of a modified energy or of auxiliary variables—e.g., in many Invariant Energy Quadratization (IEQ) and Scalar Auxiliary Variable (SAV) formulations, only a modified energy is dissipated.

Recent advances have made it possible to design schemes for a wide range of PDEs where the exact original energy dissipation property holds at the discrete level, often with high temporal accuracy and unconditional (or nearly unconditional) stability.

2. Key Families of Original-Energy-Dissipative Methods

Various algorithmic frameworks have been developed to achieve original energy dissipation:

a) Exponential and Exponential Time-Differencing Runge-Kutta (ETDRK) Schemes

High-order ETDRK schemes, particularly with rescaling post-processing, ensure the original energy is dissipated under a mild time-step condition when the nonlinearities are (locally) Lipschitz. The key construction involves a stabilized operator split and polynomial interpolation of the nonlinear source. Rescaling is introduced to preserve the maximum bound principle unconditionally, and the original energy dissipation is proved by discrete energy estimates, yielding optimal error rates and high-order performance (Quan et al., 2024). Parametric studies also exist for explicit exponential RK (EERK) schemes, where dissipation is established via conditions on differentiation matrices and the "average dissipation rate" is used for method comparison and parameter selection (Liao et al., 2024).

b) Discrete Variational Derivative (DVD) Methods

DVD methods construct a fully implicit, arbitrarily high-order Runge-Kutta time discretization by enforcing a mean-value identity for the energy increment. The resulting multi-stage schemes, when their tableau coefficients meet algebraic constraints, unconditionally dissipate the original (not modified) energy (Huang, 2022).

c) Lagrange Multiplier and Energy Correction Approaches

Lagrange multiplier techniques impose energy dissipation via scalar corrections at each time step. The semi-implicit combined Lagrange multiplier schemes compute a regular semi-implicit step, check for energy increase, and apply a minimal 1D correction (if needed) to restore the discrete energy law exactly. This enables high-order original-energy-dissipative schemes with minimal extra computational cost, suitable for phase field and general dissipative systems (Liu et al., 2023, Fang et al., 2024).

d) Improved Scalar Auxiliary Variable (iSAV) Schemes

iSAV schemes modify the original SAV algorithms by introducing stabilization and updating the auxiliary variable to ensure the discrete decay of the original energy, rather than just a modified version. These schemes are linear, efficient, and provably dissipative for a wide class of gradient flows (Chen et al., 2024).

e) Runge-Kutta–IEQ Formulations

For polynomial nonlinearities (e.g., Cahn–Hilliard), certain symplectic Runge-Kutta IEQ schemes have been shown to preserve the original energy dissipation law up to arbitrary order, provided the tableau satisfies strict symplectic conditions (Zhang et al., 2021).

f) Implicit-Explicit (IMEX) Runge-Kutta Approaches with Truncation

IMEX Runge-Kutta schemes with suitably truncated nonlinearities and stability parameters can yield unconditional original-energy-dissipativity for high-order nonlinear gradient flows (such as phase-field-crystal models) without global Lipschitz assumptions, with optimal $L^\infty$ error bounds (Li et al., 9 Jan 2026).

g) Variational and Auxiliary Variable Approaches

Schemes based on convexity and constrained optimization of the discrete energy functional, such as the modified Crank–Nicolson minimization for chemotaxis models, exact convex projection for mass/positivity, and a posteriori correction for energy dissipation, have yielded the first second-order finite-difference algorithms dissipating the original free energy for complex systems (Ding et al., 2024, Fang et al., 2024).

3. Stability, Accuracy, and Structural Preservation

a) Temporal Order and CFL-type Restrictions

• Many original-energy-dissipative schemes now attain arbitrary high temporal order ($r$), often under only a mild ($\tau$-dependent) stability restriction. For stabilized ETDRK-$r$, the threshold is explicit in terms of the stabilizer and interpolation node Vandermonde matrices (Quan et al., 2024).
• DVD and Lagrange-multiplier-based schemes often enjoy unconditional energy dissipation, with no upper bound on time-step.
• Weak restrictions (e.g., lower bounds on $\Delta t$ depending on mesh for finite difference settings) arise for positivity and dissipation in higher-order spatial discretizations (Hu et al., 2021).

b) Positivity, Maximum Bound, and Mass Conservation

• Many constructions guarantee unconditional preservation of sign (positivity) or bound (maximum principle) for key variables (e.g., densities, concentrations, order parameters), essential for blowup prevention, entropy preservation, and physical realism (Quan et al., 2024, Hu et al., 2021, Fang et al., 2024, Ding et al., 2024, Shen et al., 2020).
• Several classes ensure mass or other structural invariants are exactly discrete-conserved (Bailo et al., 2018, Fang et al., 2024).

c) Uniqueness and Convergence

• Discrete convexity or monotonicity ensures unique solvability in most schemes, even for fully nonlinear updates (e.g., via minimization of a discrete convex functional).
• Original energy dissipation allows for robust $L^\infty$ and $L^2$ error estimates up to optimal order in both space and time, without CFL constraints (Ding et al., 2024, Li et al., 9 Jan 2026).

4. Representative Applications and Testcases

Original-energy-dissipative schemes are now available and analyzed for:

• Allen–Cahn and Cahn–Hilliard equations (Ginzburg–Landau, Flory–Huggins, logarithmic potentials) (Quan et al., 2024, Zhang et al., 2021, Chen et al., 2024)
• Keller–Segel and Patlak–Keller–Segel chemotaxis models, with arbitrary mobilities and singular entropy nonlinearities (Ding et al., 2024, Fang et al., 2024, Hu et al., 2021)
• Aggregation-diffusion/Fokker–Planck/Wasserstein gradient flows: Positivity, mass, and energy dissipation are simultaneously enforced in high spatial order (Bailo et al., 2018, Hu et al., 2021)
• Poisson–Nernst–Planck (classical and modified) models in electrochemistry and ion transport, including steric interactions and nonlinear entropy (Ding et al., 2023, Shen et al., 2020)
• Navier–Stokes/Darcy (single and two-phase) flows: Enforced energy dissipation at linear solve complexity via prediction–correction and auxiliary variable relaxation (Weng et al., 8 Jun 2025, Li et al., 1 Feb 2026)
• Phase-field crystal and high-order Swift–Hohenberg type equations, using truncated and stabilized high-order IMEX RK (Li et al., 9 Jan 2026)
• Time-fractional Allen–Cahn equations: Nonlocal free energies are monotonically dissipated under general meshes (Hou et al., 2021)
• Dissipative wave and Josephson array problems: Explicit energy rate preservation for dynamical thresholds (Macías-Díaz et al., 2011)

Numerical evidence highlights monotonic decay of the original energy for all schemes, independent of time step and solution smoothness, with optimal convergence rates and preservation of mass and positivity under mesh refinement (Quan et al., 2024, Fang et al., 2024, Ding et al., 2023, Ding et al., 2024). Examples include blowup dynamics, phase transition and coarsening, pattern formation, and accurate thresholds for nonlinear supratransmission.

5. Practical Implementation and Extensions

A range of spatial discretization techniques integrate seamlessly with original-energy-dissipative time schemes, including:

• Central difference, higher-order (e.g., $Q^1$, $Q^2$, spectral) finite difference (Hu et al., 2021)
• Finite volume (cell-centered, unstructured meshes, multislope limiters) (Ding et al., 2023, Bailo et al., 2018)
• Discontinuous Galerkin methods with exact energy balance (Giesselmann et al., 2012)
• Efficient (FFT-based) solution of constant-coefficient linear systems per timestep (Weng et al., 8 Jun 2025, Hu et al., 2021, Huang, 2022)

Several methods admit decoupled or linearized solution at each step (e.g., correction for energy via scalar variable update), reducing computational cost relative to fully implicit or nonlinearized approaches (Liu et al., 2023, Li et al., 1 Feb 2026, Chen et al., 2024).

Recent works demonstrate frameworks unifying many of these methodologies under a general variational or projection-correction principle, enabling extensions to cross-diffusion, variable-mobility, nonlocal, high-order, and multiphysics systems, as well as arbitrary high-order time stepping and adaptivity (Fang et al., 2024, Huang, 2022, Li et al., 9 Jan 2026).

6. Outlook and Open Directions

While original-energy-dissipative schemes are now state-of-the-art for a large class of dissipative PDEs, several challenges and research directions remain:

• Construction and analysis of explicit or linearly-implicit high-order schemes (e.g., EERK) with provable unconditional energy stability is an active area, particularly for fourth and higher-order time discretizations (Liao et al., 2024).
• Rigorous analysis and extension to adaptive time stepping, variable coefficients, and fully discrete (space-time) error bounds, especially for coupled and multi-physics systems.
• Incorporation of further physical constraints—such as entropy inequalities, local positivity under complex boundary conditions, or positivity for systems with degenerate diffusion (Hu et al., 2021, Fang et al., 2024).
• Development of highly efficient solvers and scalable implementations for large, stiff, or singularly-perturbed systems with original energy-dissipative structure (Huang, 2022, Weng et al., 8 Jun 2025).

Original-energy-dissipative numerical methods are now a central element in the simulation of gradient flows, geometric PDEs, electrokinetics, multiphase fluid models, and beyond, enabling more robust, physically accurate, and theoretically justified computation of long-time nonlinear dynamics.