To prove that for any $n \times n$ matrix there are $n$ linearly independent eigen vectors if all the eigen values are distinct, I see that in a book it starts with the following matrix

$$ \textbf{Z}=(\alpha_2\textbf{I-A})\dots(\alpha_n\textbf{I-A})$$

where $\alpha_i; i=1,\dots,n$ are the eigen values. I did not understand the reason behind this. It looks like a trick.

Actually, the quantities $\textbf{Z$v_k$=0};k=2,3,\dots,n$ (which means $v_k;k=2,3,\dots,n$ are in the kernel space of $Z$) where $v_k$s are eigen vectors whereas $\textbf{Z$v_1$}$ is a scalar multiple of $v_1$ (which means $v_1$ is an eigen vector of $Z$). So, the matrix $\textbf{Z}$ has been chosen like this. I am trying to understand the meaning of these things.

  • $\begingroup$ what is the complete flow of proof in that book? $\endgroup$ – al-Hwarizmi Aug 25 '13 at 7:44
  • $\begingroup$ Are you sure the indices in the product start at $2$ (so there is not $\alpha_1$ in the formula)? $\endgroup$ – Marc van Leeuwen Aug 25 '13 at 9:13

For every eigenvalue $\alpha$, the matrix $\alpha I - A$ contains the respective eigenvectors of $A$ in its kernel:

$$ Av = \alpha v \qquad \Longleftrightarrow \qquad (\alpha I - A) v = 0 \ . $$

So, I guess your book could proceed like this: let $v_1, \dots , v_n$ be eigenvectors of $\alpha_1, \dots , \alpha_n$ eigenvalues. Assume there is a linear combination such that

$$ \lambda_1 v_1 + \dots + \lambda_n v_n = 0 \ . $$

Applying matrix $Z$ to both sides of this equality, we get

$$ \lambda_1 (\alpha_2 - \alpha_1) \dots (\alpha_n - \alpha_1) v_1 = 0 \ . $$

Since $v_1 \neq 0$, by definition of eigenvector, and all the eigenvalues $\alpha_i$ are different, it follows that

$$ \lambda_1 = 0 \ . $$

Applying induction, or changing matrix $Z$ suitably, you would also get that all coeffitients $\lambda_i$ are necessarily zero. Hence vectors $v_1, \dots , v_n$ are linearly independent.

  • $\begingroup$ Roig: yes, this is the solution. I was trying to understand the intuition behind this trick of choosing $Z$ using the fact that it is some kind of linear map. $\endgroup$ – RIchard Williams Aug 25 '13 at 9:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.