Now . Suppose , and normalize the orthogonal good and bad components:
Their disjoint computational supports make an orthonormal basis. For , the two-dimensional matrix for phase amplitude amplification is
To obtain its first column, the marked phase first multiplies by , then adds its projection onto ; the second column is obtained similarly without . This also proves unitarity: the rank-one phase has eigenvalues on this plane, and the marked phase has eigenvalues , all of unit modulus. Their negative product is unitary for any real phases.
Using the unnormalized pair instead would produce a differently scaled matrix, which is not in general unitary in the ordinary Euclidean metric. At or one component vanishes and the relevant search space is one-dimensional.