ewcommand{\b}{\v b} \end{equation*} Otherwise, we must find the ’next most-orthogonal’ face \(j_2\), with the restriction that it must not be ‘behind’ the previous face: In the \(2\) dimensional cases it could be that there are many lines very close to orthogonal to \(\t\) on one side of the optimal solution, and none on the other side. Then ideally we can repeatedly find the most-orthogonal face and once we find \(n\) of them (or once \(\t\) is a linear combination of \(A_{j_1}\) to \(A_{j_k}\)) we know that the optimal solution is at the intersection of those \(n\) faces.