Suppose we already have m linearly independent eigenvectors $\mu_1, \mu_2, ... \mu_m$. Say $A\mu_i = \lambda\mu_i$ for $i = 1, 2, ..., m$ where $\lambda_1, \lambda_2, ..., \lambda_m$ are not necessarily distinct. For a new eigenvalue $\mu(\mu \ne \lambda_i \;~for~\; i = 1, 2, ..., m)$ of A, let $\{ v_1, v_2, ..., v_p\}$ be a basis for the eigenspace $E_{\mu}$.

Prove that $\{ \mu_1, \mu_2 ..., \mu_m, v_1, v_2, ..., v_p\}$ is linearly independent. (Hint: consider the vector equation

$a_1\mu_1 + a_2\mu_2 + ... + a_m\mu_m + b_1v_1 + b_2v_2 + ... + b_pv_p = 0$ --- (1)

By using the property of eigenvectors, show that

$a_1(\lambda_1-\mu)\mu_1 + a_2(\lambda_2-\mu)\mu_2 + ... + a_m(\lambda_m-\mu)\mu_m = 0$ ---(2)

I wonder how proving equation(2) in Hint help proving equation(1) in which $a_1, a_2, ..., a_m, b_1, b_2, ..., b_p$ are all 0?

  • $\begingroup$ Perhaps you should not use $\mu$ to denote both eigenvectors and eigenvalues (just a suggestion). $\endgroup$ Apr 26 '14 at 7:34
  • $\begingroup$ I just copy the Question in book. In the book, it is using that notation, so I didn't change anything. should I change some notation to re-formulize this question? $\endgroup$ Apr 26 '14 at 9:47

If you can prove equation (2), then you basically prove that $a_1=\cdots=a_m=0$.


Since $\{\mu_1,\ldots,\mu_m\}$ is linearly independent we must have $a_i(\lambda_i-\mu)=0$ for each $1\leq i \leq m$, and since $\lambda_i$ is distinct from $\mu$, we have $\lambda_i-\mu \neq 0$...it follows that $a_i=0$. This reduces equation (1) to $b_1v_1 + b_2v_2 + ... + b_pv_p = 0$ ... you can take it from here.


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.