why the derivative does not exist at a cusp