Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Numerical analysis for saddle dynamics of some semilinear elliptic problems

Lei Zhang, Xiangcheng Zheng, Shangqin Zhu

TL;DR

This work develops a continuous-in-space index-1 saddle dynamics framework to compute transition states of semilinear elliptic PDEs by recasting the problem as a parabolic system whose stationary points coincide with saddle points. It provides a rigorous numerical analysis for both semi-discrete and fully discrete finite element schemes, proving well-posedness, $H^1$ stability, and error estimates, while addressing gradient nonlinearity and a norm-constraint via a normalization projection. A key contribution is the demonstration of index preservation under discretization, supported by a careful error-splitting approach and auxiliary estimates, complemented by a fully discrete scheme that remains robust through normalization. Numerical experiments in one and two dimensions validate the theoretical results, showing correct convergence rates and distinctive index-1 saddles arising from different initial guesses. The paper lays a foundation for extending saddle dynamics to higher-index problems and more complex PDE settings, with potential implications for reliably uncovering transition states in nonlinear systems.

Abstract

This work presents a numerical analysis of computing transition states of semilinear elliptic partial differential equations (PDEs) via the index-1 saddle dynamics, or equivalently, the gentlest ascent dynamics. To establish clear connections between saddle dynamics and numerical methods of PDEs, as well as improving their compatibility, we first propose the continuous-in-space formulation of saddle dynamics for semilinear elliptic problems. This formulation yields a parabolic system that converges to saddle points. We then analyze the well-posedness, $H^1$ stability and error estimates of semi- and fully-discrete finite element schemes. Significant efforts are devoted to addressing the coupling, gradient nonlinearity, nonlocality of the proposed parabolic system, and the impacts of retraction due to the norm constraint. The error estimate results demonstrate the accuracy and index-preservation of the discrete schemes.

Numerical analysis for saddle dynamics of some semilinear elliptic problems

TL;DR

This work develops a continuous-in-space index-1 saddle dynamics framework to compute transition states of semilinear elliptic PDEs by recasting the problem as a parabolic system whose stationary points coincide with saddle points. It provides a rigorous numerical analysis for both semi-discrete and fully discrete finite element schemes, proving well-posedness, stability, and error estimates, while addressing gradient nonlinearity and a norm-constraint via a normalization projection. A key contribution is the demonstration of index preservation under discretization, supported by a careful error-splitting approach and auxiliary estimates, complemented by a fully discrete scheme that remains robust through normalization. Numerical experiments in one and two dimensions validate the theoretical results, showing correct convergence rates and distinctive index-1 saddles arising from different initial guesses. The paper lays a foundation for extending saddle dynamics to higher-index problems and more complex PDE settings, with potential implications for reliably uncovering transition states in nonlinear systems.

Abstract

This work presents a numerical analysis of computing transition states of semilinear elliptic partial differential equations (PDEs) via the index-1 saddle dynamics, or equivalently, the gentlest ascent dynamics. To establish clear connections between saddle dynamics and numerical methods of PDEs, as well as improving their compatibility, we first propose the continuous-in-space formulation of saddle dynamics for semilinear elliptic problems. This formulation yields a parabolic system that converges to saddle points. We then analyze the well-posedness, stability and error estimates of semi- and fully-discrete finite element schemes. Significant efforts are devoted to addressing the coupling, gradient nonlinearity, nonlocality of the proposed parabolic system, and the impacts of retraction due to the norm constraint. The error estimate results demonstrate the accuracy and index-preservation of the discrete schemes.

Paper Structure

This paper contains 14 sections, 9 theorems, 99 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Mathematical analysis
  3. Analysis of semi-discrete scheme
  4. Numerical scheme
  5. Well-posedness issue
  6. Error estimates
  7. Analysis of full discretization
  8. Numerical scheme
  9. Well-posedness issue
  10. Analysis of normalized projection
  11. Auxiliary estimates
  12. Error estimates
  13. Numerical experiments
  14. Concluding remarks

Key Result

Theorem 1

\newlabelthm210 Assume that the points $(u^*,{v^*_1})\subset H^2(\Omega)\cap H^1_0(\Omega)$ satisfy $\|v^*_1\|=1$ and the operator $-\Delta-f'(u^*)$ under zero Dirichlet boundary conditions has non-zero eigenvalues $\lambda_1^*\leq \lambda_2^*\leq \lambda_{3}^*\leq \cdots$. Then the following two

Figures (2)

  • Figure 1: Plots of $u_h(T)$ and $F(u_h(T))$ for Example 1.
  • Figure 2: Plots of $u_h(T)$ and $F(u_h(T))$ for Example 2.

Theorems & Definitions (23)

  • Theorem 1
  • Proof 1
  • Remark 2.1
  • Remark 3.1
  • Lemma 1
  • Proof 2
  • Theorem 2
  • Proof 3
  • Theorem 3
  • Proof 4
  • ...and 13 more