Abstract
We compare six different categories of numerical methods for the evaluation of functions of the matrix exponential. These functions are required for exponential integrators, and are not straightforward to evaluate because they are highly susceptible to rounding errors when the matrix has small eigenvalues. The comparison takes into account both accuracy and computational time. A scaling and squaring algorithm and a diagonalisation algorithm are both found to be efficient.
Original language | English (US) |
---|---|
Pages (from-to) | 468-486 |
Number of pages | 19 |
Journal | Applied Numerical Mathematics |
Volume | 59 |
Issue number | 3-4 |
DOIs | |
State | Published - Mar 2009 |
Externally published | Yes |
All Science Journal Classification (ASJC) codes
- Numerical Analysis
- Computational Mathematics
- Applied Mathematics
Keywords
- Exponential integrators
- Exponential time differencing methods
- Matrix exponential
- Stiff differential equations