I wish to prove that any $2\times 2$ matrix $T$ with only one eigenvalue $ \lambda$ of geometric multiplicity 1 is similar to one of the form

$\left[\begin{array}{cc} \lambda & 1 \\ 0 &\lambda \end{array}\right]$.

Now if I choose my basis to be $v$, where $v$ is some eigenvector of $\lambda$ and any vector $w$ not in null $(T - \lambda I)$ this should work. It is obvious in this basis how the first column of the matrix above comes about,

and for the second column, we note that

$T(w) = (T-\lambda I)w + \lambda w$ and by Cayley-Hamilton it is clear that $(T - \lambda I)w$ is an eigenvector of $T$. However this only means that in the matrix above I have a constant, not necessarily 1 in the slot where it is sitting.

How can I choose $w$ so that $(T-\lambda I)w$ is an eigenvector of $T$ of eigenvalue 1?



Once you have $\left[\begin{array}{cc} \lambda & c \\ 0 &\lambda \end{array}\right]$ with $c\neq0$, you can use


  • 1
    $\begingroup$ You're right, so my basis instead of being $\{v,w\}$ would now be $\{cv,w \}$ for $c\neq 0$. $\endgroup$ – user38268 Oct 13 '11 at 6:27
  • $\begingroup$ @Benjamin: Whether a vector is an eigenvector of an operator is basis-independent. I didn't look at what you wrote about eigenvectors in detail; I understood your problem to be that you could reach the form $\left[\begin{array}{cc} \lambda & c \\ 0 &\lambda \end{array}\right]$ but not $\left[\begin{array}{cc} \lambda & 1 \\ 0 &\lambda \end{array}\right]$, so that's what I addressed. $\endgroup$ – joriki Oct 13 '11 at 7:21

Your Answer

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