Let $\dim\ker(\lambda I-A)=1$, and $v_1\in\ker(\lambda I-A)$; $v_1\neq0$.

Also, we know that, then, $\dim\ker(\lambda I-A)^2=2$. Take $v_2\in \ker(\lambda I-A)^2\setminus\ker(\lambda I-A)$; $v_n\neq 0$.

Why can one say that $Av_2=v_1+\lambda v_2$?

I know that $Av_2\in\ker(\lambda I-A)^2$, then there are $\alpha$ and $\beta$ s.t. $$Av_2=\alpha v_1+\beta v_2$$ bu how can one conclude that $\alpha=1$ and $\beta=\lambda$?

I'm trying to understand Jordan Normal Form. Because we know that in this case $$J=\left(\begin{array}{cc} \lambda & 1\\ 0 & \lambda\end{array}\right),$$ in a way that $\{v_1,v_2\}$ is its Jordan basis.


You have $$ 0=(\lambda I-A)^2v_2=(\lambda I-A)\,(\lambda I-A)v_2. $$ This shows that $(\lambda I-A)v_2\in\ker(\lambda I-A)$. Since $\dim\ker(\lambda I-A)=1$, any vector in it is a scalar multiple of $v_1$. Thus $$ (\lambda I-A)v_2=\alpha\,v_1 $$ for some scalar $\alpha$. This we can rewrite as $$ Av_2= - \alpha v_1 + \lambda v_2. $$ Note that the constant in front of $v_1$ cannot be avoided if you are taking $v_1$ as any element of $\ker(\lambda I-A)$ and $v_2$ as any element of $\ker (\lambda I-A)^2$.

Edit: Carl's comment.

Now, if we allow ourselves to choose $v_2$, then we can replace the $v_2$ above by $v_2/(-\alpha)$, and then the formula becomes $$ Av_2=v_1+\lambda v_2. $$ Note that $\alpha\ne0$, because otherwise $v_2$ would be an eigenvector for $\lambda$, which it is not by choice.

| cite | improve this answer | |
  • 2
    $\begingroup$ As to why $\alpha=1$, first of all note that if $\alpha=0$, then $v_2$ would be an eigenvector itself (which it isn't). Hence, you can divide both sides of the equation by $-\alpha$ to get $$A \cdot {v_2 \over -\alpha} = v_1 + \lambda \cdot {v_2 \over -\alpha},$$ so if we use $\displaystyle v_2'= {v_2 \over -\alpha}$ instead of $v_2$, then $$Av_2' = v_1 + \lambda v_2'.$$ $\endgroup$ – Christopher Carl Heckman Apr 17 '16 at 2:38
  • 1
    $\begingroup$ And thanks for the shorter solution! I was about to post an algebraic monstrosity ... $\endgroup$ – Christopher Carl Heckman Apr 17 '16 at 2:39
  • $\begingroup$ Thanks, Carl. I incorporated your comment in the answer. $\endgroup$ – Martin Argerami Apr 17 '16 at 4:52

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.