MA2.101 - Linear Algebra | Linear Algebra Lecture 18

with Prof. Girish Varma, Prof. Indranil
Apr 03, 2021 - Saturday
Written by: Pratyaksh Gautam

More on eigenvalues and eigenvectors

Eigenvectors corresponding to distinct eigenvalues are linearly independent.

Suppose $M$ has $n$ linearly independent eigenvectors,${ v_1, v_2, … v_n}$, with corresponding eigenvalues ${ \lambda_1, \lambda_2, … \lambda_n}$

Suppose \(w = \alpha_1 v_1 + \alpha_2 v_2 + ... + \alpha_n v_n\) \(\implies Mw = \alpha_1 \lambda_1 v_1 + \alpha_2 \lambda_2 v_2 + ... + \alpha_n \lambda_n v_n\)

In this representation $(\alpha_1, … \alpha_n)$ are co-ordinates of $W$ in ${v_1, … v_n}$. If we change basis from ${ e_1, e_2, … e_n}$ to ${ v_1, v_2, … v_n }$, then the linear transformation corresponding to $M$ acts like a diagonal matrix.

So in order to change our basis, we must premultiply by a change of basis matrix, say $P$, where $P$ converts from the new basis to our old standard basis. Thus, if in the old basis, we had $w \mapsto Mw$, then in our new basis, we have $\bar w = P^{-1}w$, and the new mapping $\bar w \mapsto P^{-1}MP \bar w$

So if we wished to think of this linear transform in our new basis only, as some transform $\bar M$, such that $\bar w \mapsto \bar M \bar w$, then clearly we have $\bar M = P^{-1}MP$.

Since application happens from right to left, we can think of what happens to $\bar w$ sequentially, for a more intuitive understanding.

  1. $\bar w$ gets converted back to $w$, since $w = P \bar w$. Now we’re in the old standard basis.
  2. $M$ gets applied to $w$, the original transform that we knew, and this operation happens in the old basis.
  3. $P^{-1}$ finally gets applied to this result as well, giving us the result in our new “eigenbasis”.

Note that the eigenvalues of $\bar M$ will be given directly by the diagonal elements of the matrix, which are simply all the $\lambda_i$’s, i.e the same as the eigenvalues of of $\bar M$

Equivalence relation

Consider a relation for $n \times n$ matrix defined as such

$M \equiv \bar M$ iff there exists invertible $P$ such that $P^{-1}MP = \bar M$.

$M$ is said to be diagonalizable iff $M \equiv$ to a diagonal matrix.

Diagonalizable matrices have some useful properties which stem from the diagonal matrices they are equivalent to.

\(M^tv\) \(= (PDP^{-1})^tv\) \(= (PD^tP^{-1})v\)

And computing $D^t$ itself is trivial given $D$ is a diagonal matrix; we simply take the elements along the diagonal to the power $t$ for each one.