A matrix is orthogonal if
This means that the columns of are orthogonal to each other.
Notation: : transpose conjugate
Unitary: same for complex matrices: .
Orthogonal matrices are very important in numerical linear algebra. They often guarantee that an algorithm will be accurate.
Orthogonal matrices are isometries, that is, they conserve the 2-norm of a vector:
If is square, then its inverse is . In that case, can be decomposed into a sequence of rotations and reflections.
Fun fact: Cartan–Dieudonné theorem. Every orthogonal transformation in can be described as the composition of at most reflections.
The matrix is not equal to if is rectangular (i.e., ).
The matrix represents an orthogonal projection onto along .
Dot product, Vector norms, Operator and matrix norms, Projection