next up previous contents
Next: Smoothing Up: THE FOURIER METHOD Previous: Epilogue

Skew-Symmetric Differencing



There are two main approaches to enforce stability at this point: skew-symmetric differencing and smoothing. We discuss these issues in the next two subsections.

The essential argument of well-posedness for symmetric hyperbolic systems with constant coefficients is the fact that (say in the 1-D case) is a skew-adjoint operator. With variable coefficients this is also true, modulo low-order bounded terms, i.e.,
The stability proofs of spectral methods follow the same line, i.e., we have in the Fourier space, compare (meth_spec.45),
and stability amounts to show that the second term in (skew.2) is bounded: for then we have in (skew.2) (as in ((skew.1) ) a skew-adjoint term with an additional bounded operator. The difficulty with the stability of pseudo-spectral methods arises from the fact that the second term on the right of (skew.2) is unbounded,
To overcome this difficulty, we can discretized the symmetric hyperbolic system (again, say the 1-D case)
when the spatial operator is already put in the ``right'' skew-adjoint form, compare (skew.1),

The pseudospectral approximation takes the form
In the Fourier space, this gives us
Now, is symmetric because is, is bounded and stability follows.

Eitan Tadmor
Thu Jan 22 19:07:34 PST 1998