[jollage]

Hey, I am not sure if it's the right place to post a question about the generation of derivative matrix in implementation of spectral method. I only have one basic question:

In forming the derivative matrices D1, D2, D3 and so forth with the Dirichlet boundary condition at the end points of the Chebyshev points, namely -1 and 1, the conventional way is to remove the first and last rows and first and last columns of the derivative matrices, and also remove the end points of the Chebyshev points. This makes sense because the values at the ends points are zeros.

My question is why couldn't we retain these rows and columns and also retain the end points of the Chebyshev points(sure, the values are 0 there.)? I tried this, but it seem not work.

Does anyone have a clue?