Abstract
This paper considers the stability of one-leg linear multistep methods applied to linear nonautonomous systems of equations, in which the governing matrix is nonsymmetric and is subject to a constant rotation. It is shown that the stability of the underlying system and of the one--leg methods may be analysed using eigenvalues, following a transformation to a rotating frame. For 2-dimensional systems, general conditions for numerical instability are derived. Stable systems are constructed that are unstable for the Backward Euler and BDF2 methods. A family of neutrally stable systems is identified that is numerically unstable for the BDF3 and BDF4 methods, for all sufficiently small step sizes.
Original language | English |
---|---|
Pages (from-to) | 103-112 |
Number of pages | 10 |
Journal | JNAIAM. Journal of Numerical Analysis, Industrial and Applied Mathematics |
Volume | 4 |
Issue number | 1-2 |
Publication status | Published - 2009 |