A linear point transformation y = ax viewed as occurring in three steps
Let A be an nxn matrix all of whose eigenvalues are distinct.. Then A is similar to a diagonal
matrix D whose diagonal elements are these eigenvalues. In addition, all of the n eigenvectors
are linearly independent. Let
be the normalized eigenvectors and let them
constitute a basis for an Eigenvector Coordinate System (an oblique coordinate system centered
at the origin called the “eigenbasis”). Let Z be a matrix whose columns contain the normalized
eigenvectors
. Let vector X represent the coordinates of a point expressed
relative to the usual E-basis and let vector XZ represent the coordinates of the same point
expressed relative to the Z-basis (eigenbasis). Then
X = ZXZ
and
XZ = Z -1 X
The linear point transformation effected by matrix A is that same one effected by the diagonal matrix D in the Eigenvector Coordinate System. It is just viewed from a different coordinate system. The linear point transformation effected by matrix A can be viewed as occurring in three steps:
1. A change of basis from the usual coordinate system (E-basis) to the Eigenvector Coordinate System (the canonical coordinate system where the actual point transformation is performed). This change of basis corresponds to the coordinate transformation XZ = Z -1 X where the vector X is expressed relative to the Z-basis (eigenbasis) as XZ .
2. A linear point transformation effected by the diagonal matrix D in the Eigenvector Coordinate
System. This point transformation is given by U = DXZ . The effect of this transformation is
simply stretching (or compressing) effects directed in the directions of the different coordinate
system axes
with magnitudes given by the eigenvalues
.
3. A change back to to the original E-basis given by Y = ZU .
Putting these three steps together we have
Y = ZD Z -1 X
thus
A = ZD Z -1