Computational enhancement of discrete gradient method

Artur Kobus

DOI: https://doi.org/10.48550/arXiv.1912.07253

2020-02-16

Abstract:We propose new numerical approach to non-conservative dynamical systems. Our method being of low order, enhances qualitative performance of standard discrete gradient algorithm, thank to new concept of a reservoir. Paper is of explanatory character, focusing on concrete non-linear physical systems. Superiority of our new method with respect to standard discrete gradient method is observed.

Numerical Analysis,Computational Physics

What problem does this paper attempt to address?

This paper attempts to solve the numerical simulation problem of non - conservative dynamical systems. Specifically, the author proposes a new numerical method to improve the discrete gradient method, making it have better qualitative performance when dealing with non - conservative systems.

### Main problems and background of the paper

1. **Limitations of traditional methods**:
- For solving differential equations, there are usually two main methods: high - order general integrators (such as Runge - Kutta methods, multi - step methods, etc.) and geometric integration methods.
- Geometric integration methods can accurately preserve certain qualitative characteristics of the system, such as time - reversal symmetry, symplectic structure or constants of motion.
- However, the traditional discrete gradient method is mainly used for conservative systems (such as Hamiltonian systems, Poisson systems and Lyapunov - type systems), and it has poor performance for non - conservative systems.

2. **Introduction of new concepts**:
- The author proposes a new concept - "computational invariant of motion", by introducing a new variable \(z\) called "reservoir variable", so that an approximately conserved quantity \(K\) can be constructed even in non - conservative systems.
- This new method is especially suitable for nonlinear, non - conservative dynamical systems and can improve the qualitative performance of the standard discrete gradient algorithm.

### Specific problem description

- **Target differential equation**:
Consider an initial value problem (IVP):
\[
\dot{\mathbf{x}}=f(\mathbf{x}),\quad \mathbf{x}(t_0)=\mathbf{x}_0
\]
where \(\mathbf{x}\) is the state vector and \(f\) is the dynamic function of the non - conservative system.

- **Core idea of the new method**:
- Introduce the reservoir variable \(z\), defined as:
\[
z = -\int D(x,y)y\,dt
\]
- Construct a new conserved quantity \(K\):
\[
K = H+z
\]
where \(H\) is the energy function of the system (for a simple harmonic oscillator \(H=\frac{1}{2}x^2+\frac{1}{2}y^2\)).
- The new numerical scheme ensures that \(K\) remains conserved during the discretization process.

### Experimental verification

The author carried out numerical experiments through several specific nonlinear systems (such as damped harmonic oscillators, Van der Pol oscillators and Duffing oscillators) to verify the effectiveness and superiority of the new method. The experimental results show that the new method is superior to the standard discrete gradient method and other low - order integrators in maintaining the energy - decreasing characteristics and trajectory stability.

### Conclusions and prospects

1. **Current progress**:
- The new method shows significant advantages in dealing with non - conservative systems, especially in terms of the importance of energy - decreasing characteristics.
- The discrete gradient process can be used even if the system has no explicit constants of motion or Lyapunov functions.

2. **Future research directions**:
- Further theoretical analysis of the properties of the new method.
- Explore better intermediate point selection strategies \(x_i^*,y_i^*\) and long - time performance tests.
- Extend to higher - dimensional systems and attempt matrix - form implementation.

In conclusion, this paper aims to improve the application of the discrete gradient method in non - conservative dynamical systems by introducing the concepts of reservoir variable and computational invariant, thereby providing a more effective numerical simulation tool.