Source

Adapted from my own handwritten and hand-typed notes from Year 1 to Year 3 of my EEE degree

Diagonalisation rewrites a square matrix as with diagonal. In the eigenvector "coordinates" the matrix just scales each direction, which makes powers and differential equations easy.

Theorem

An matrix is diagonalisable if and only if it has linearly independent eigenvectors. Then

The columns of are the eigenvectors, and the diagonal of holds the matching eigenvalues in the same order.

Why It Works

Stack the equations side by side:

so . If the eigenvectors are independent, is invertible and .

Method

Steps:

  1. Find all eigenvalues and eigenvectors
  2. Check you have independent eigenvectors. If not, is not diagonalisable
  3. Build from the eigenvectors and from the eigenvalues in matching order

Worked Example

For , the eigenpairs are , and , . So

Not Diagonalisable Example

has eigenvalues . The double eigenvalue has only one eigenvector , so there are only two independent eigenvectors in total. is not diagonalisable.

Sufficient condition: distinct eigenvalues always give independent eigenvectors.

Powers of A

The same trick gives , used for Systems of Linear Differential Equations.