-
For ,
forward Euler in time with centered in space finite differencing
(second-order) led to time step condition .
This is computationally prohibitive, leading to small time steps
as
-
Try an implicit-in-time scheme. Since spatial discretization is
second-order in space try trapezoid which is second-order in time
-
This is Crank-Nicolson, requiring solution of a linear system at
each time step
Since linear system is tridiagonal, computational effort is small
per time step
-
-
Stability criterion:
with
within region of stability of numerical scheme
-
Since
is discretization of ,
eigenvalues of
related to those of
-
Example: Forward Euler in time with time step , centered
in space with step
-
Eigenvalues of ,
guess eigenvector is discretization of
-
Eigenvalues of
are