next up previous contents
Next: Flux Integrals Up: Elemental Spatial Operators Previous: Elemental Spatial Operators

Integration

We take advantage of the tensor product element coordinate systems to perform integration. The integrations over each element can be performed as a set of one-dimensional integrals using Gauss quadrature. If we used the reference coordinate systems this would be very difficult since the limits of the ``collapsed'' elements are not constant.

We first describe the choice of quadrature type for integrating each direction. We will then motivate the inclusion of quadrature with non-constant weights in order to reduce the number of points we use.

In two-dimensions we consider integrals of the form :

In three-dimensions:

We use the Gauss weights that will perform the discrete integral of a function as a sum:

\begin{displaymath}\int_{-1}^{1} (1-z)^{\alpha}(1+z)^{\beta}f(z) dz = \sum_{i=0}^{N-1} f(z_i^{\alpha,\beta})w_i^{\alpha,\beta}\end{displaymath}
This will be used in each of the d directions in the d-dimensional elements. In table 4.1 we show the type of Gaussian quadrature we use in each of the `a',`b' and `c' directions.
 
 
Table 4.1: GLL implies Gauss Lobatto Legendre which is the Gauss quadrature for a constant weight function with both $x=\pm 1$ points endpoints included. $GRJ_{\alpha,\beta}$ implies Gauss Radau Jacobi quadrature with $(\alpha,\beta)$ weights and the endpoint x=-1 included.
Element `a' `b' `c'
Triangle GLL GRJ1,0 -
Quadrilateral GLL GLL -
Tetrahedron GLL GRJ1,0 GRJ2,0
Pyramid GLL GLL GRJ2,0
Prism GLL GLL GRJ1,0
Hexahedron GLL GLL GLL


next up previous contents
Next: Flux Integrals Up: Elemental Spatial Operators Previous: Elemental Spatial Operators 
T. Warburton

10/24/1998