It may be a quite silly question but I am having trouble with this.

My question is that if a nonzero vector $v\in V$ is a generalized eigenvector for a linear operator $T: V\to V$ such that $(T-\lambda_1)^{d_1}v=0$ and $(T-\lambda_2)^{d_2}v=0$ where $d_1$ and $d_2$ are positive integers, is it necessarily true that $\lambda_1 = \lambda_2$?

For example, suppose T is a nilpotent linear operator. Then every vector is a generalized eigenvector with eigenvalue 0. Isn't there any generalized eigenvector with nonzero eigenvalue?

  • 1
    $\begingroup$ A important result in Jordan-normal-form theory is $$ V=R_{\lambda_1}\left(\mathscr{A}\right)\oplus R_{\lambda_2}\left(\mathscr{A}\right)\oplus\cdots\oplus R_{\lambda_l}\left(\mathscr{A}\right) $$ where $R_{\lambda_i}\left(\mathscr{A}\right)=\mathrm{Ker}\left(\mathscr{A}-\lambda_i\mathrm{id}\right)^{k_i}$. If you have got here, the answer of your problem is obvious. Maybe refer to the textbook again is more helpful. $\endgroup$ – user823011 Sep 17 at 10:03

Yes, it is necessarily true that $\lambda_1 = \lambda_2$. In particular: suppose that $(T - \lambda_1)^{d_1} v = 0$ and $(T - \lambda_1)^{d_1 - 1}v \neq 0$. Then $w = (T - \lambda_1)^{d_1 - 1} v$ is non-zero and satisfies $Tw = \lambda_1 w$.

It follows that if $\lambda_2 \neq \lambda_1$, we have $$ \begin{align} (T - \lambda_1)^{d_1 - 1}[(T - \lambda_2)^{d_2}v] &= (T - \lambda_2)^{d_2}[(T - \lambda_1)^{d_1 - 1}v] \\ & = (T - \lambda_2)^{d_2} w = (\lambda_1 - \lambda_2)^{d_2}w \neq 0. \end{align} $$ It follows that $(T - \lambda_2)^{d_2}v \neq 0$.

| cite | improve this answer | |
  • $\begingroup$ Thank you for your clear proof $\endgroup$ – Henry Choi Sep 17 at 10:10

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.