The idea here is to keep the speed of a 1-d tridiagonal solver when solving 2-D or 3-D problems. The procedure is to split the laPlacian operator into x and y derivative parts. Then solve for values in 1 direction while holding values in the other direction.
More specifically:
Like SOR, it works even faster if you overcorrect. r is the overcorrection factor: 1 <= r <= 2.
General formula:
Same formula written in tridiagonal form:
Back to ATM 150 homepage