High-order linearly implicit two-step peer -- finite element methods for time-dependent PDEs

Linearly-implicit two-step peer methods are successfully applied in the numerical solution of ordinary differential and differential-algebraic equations. One of their strengths is that even high-order methods do not show order reduction in computations for stiff problems. With this property, peer methods commend themselves as time-stepping schemes in finite element calculations for time-dependent partial differential equations (PDEs). We have included a class of linearly-implicit two-step peer methods in the finite element software Kardos. There PDEs are solved following the Rothe method, i.e. first discretised in time, leading to linear elliptic problems in each stage of the peer method. We describe the construction of the methods and how they fit into the finite element framework. We also discuss the starting procedure of the two-step scheme and questions of local temporal error control. The implementation is tested for two-step peer methods of orders three to five on a selection of PDE test problems on fixed spatial grids. No order reduction is observed and the two-step methods are more efficient, at least competitive, in comparison with the linearly implicit one-step methods provided in Kardos.

[1]  C. Lubich,et al.  Linearly implicit time discretization of non-linear parabolic equations , 1995 .

[2]  Helmut Podhaisky,et al.  Multi-Implicit Peer Two-Step W-Methods for Parallel Time Integration , 2005 .

[3]  Jochen Fröhlich,et al.  Two-dimensional cascadic finite element computations of combustion problems , 1998 .

[4]  Jens Lang,et al.  ROS3P—An Accurate Third-Order Rosenbrock Solver Designed for Parabolic Problems , 2000 .

[5]  B. A. Schmitt,et al.  Numerical experiments with parallel two‐step W‐methods for DAEs , 2002 .

[6]  J. E. Pearson.  Complex Patterns in a Simple System , 1993, Science.

[7]  A. Wouwer,et al.  Adaptive Method of Lines , 2001 .

[8]  C. Lubich,et al.  Runge-Kutta approximation of quasi-linear parabolic equations , 1995 .

[9]  Gerd Steinebach,et al.  Order-reduction of ROW-methods for DAEs and method of lines applications , 1995 .

[10]  J. Verwer,et al.  Numerical solution of time-dependent advection-diffusion-reaction equations , 2003 .

[11]  E. Hairer,et al.  Solving Ordinary Differential Equations II: Stiff and Differential-Algebraic Problems , 2010 .

[12]  Helmut Podhaisky,et al.  Two-step W-methods and their application to MOL-systems , 2003 .

[13]  J. Butcher.  Numerical methods for ordinary differential equations , 2003 .

[14]  J. Lang,et al.  Two–dimensional adaptive simulation of dopant diffusion in silicon , 2001 .

[15]  Helmut Podhaisky,et al.  Rosenbrock-type 'Peer' two-step methods , 2005 .

[16]  F. Krogh,et al.  Solving Ordinary Differential Equations , 2019, Programming for Computations - Python.

[17]  M. Seebass,et al.  Impact of nonlinear heat transfer on temperature control in regional hyperthermia , 1999, IEEE Transactions on Biomedical Engineering.

[18]  Rüdiger Weiner,et al.  Parallel Two-Step W-Methods with Peer Variables , 2004, SIAM J. Numer. Anal..

[19]  Jens Lang,et al.  Adaptive Multilevel Solution of Nonlinear Parabolic PDE Systems - Theory, Algorithm, and Applications , 2001, Lecture Notes in Computational Science and Engineering.

[20]  Jens Lang,et al.  KARDOS - User"s Guide , 2002 .