I'm self-studying linear algebra and encountered a problem that I need help with. Here is the problem statement

Suppose $V$ is a finite-dimensional $\mathbb{C}$-vector space and $T\in\mathcal{L}(V)$ has $\lambda$ as an eigenvalue. Suppose also that $(z-\lambda)^m$ is a factor of the characteristic polynomial of $T$ and $(z-\lambda)^k$ is a factor of the minimal polynomial (and no larger powers of $z-\lambda$ in either case). If $\bar{v}$ is a Jordan basis for $T_{|G(\lambda,T)}$, find $\dim E(\lambda,T_{|G(\lambda,T)})$ and compute the corresponding matrix $M(T_{|G(\lambda,T)},\bar{v})$ when 1. $k=1$ and 2. $k=m$.

Notation wise, $E(\lambda,T)$ is the eigenspace of linear function $T$ with respect to eigenvalue $\lambda$ and $G(\lambda,T)$ the generalized eigenspace. $T_{|A}$ is the operator $T$ restricted to the set $A$ (i.e. the domain is set $A$).

I'm wondering if it is correct to state that $$ E(\lambda,T_{|G(\lambda,T)}) = E(\lambda,T) $$ as $E(\lambda,T) \subset G(\lambda,T)$. Also, in the case of $k=1$, I understand that the largest Jordan block is $1\times 1$. Am I correct in conjecturing that the corresponding matrix in this case is $\lambda I$? Any clue to solving this question is much appreciated :-)


1 Answer 1


Let us find all eigenvectors of $E(\lambda, T_{|G(\lambda, T)})$. Clearly, $E(\lambda, T) \subset G(\lambda, T)$ is a set of eigenvectors, and by the definition of a generalized eigenvector, there are no more eigenvectors. Thus, $$E(\lambda, T_{|G(\lambda, T)}) = E(\lambda, T)$$

We can see that $T- \lambda\mathrm{id}$ is a nilpotent map on $G(\lambda, T)$, and $k$ is the index. So

$$k=1 \implies T-\lambda \mathrm{id} = \mathbf{0}$$

which implies that every vector is an eigenvector, or $G(\lambda, T) = E(\lambda, T)$. The corresponding matrix is $\lambda \mathrm{id}$. Now

$$k = m \implies (T-\lambda \mathrm{id})^m = \mathbf{0}, \: (T-\lambda \mathrm{id})^{m-1} \neq \mathbf{0}$$ Thus there exists basis vectors $v_1, \ldots, v_m$ so that $T - \lambda \mathrm{id}$ maps

$$v_1 \mapsto v_2 \mapsto \ldots v_m \mapsto \mathbf{0}$$ If $E(\lambda, T)$ is one dimensional, then the matrix looks like this:

$$ \begin{bmatrix} \lambda & 1 & 0 & \ldots & 0 & 0 \\ 0 & \lambda & 1 & \ldots & 0 & 0 \\ 0 & 0 & \lambda & \ldots & 0 & 0 \\ \vdots & \vdots & \vdots & \ddots & \vdots & \vdots \\ 0 & 0 & 0 & \ldots & \lambda & 1 \\ 0 & 0 & 0 & \ldots & 0 & \lambda \\ \end{bmatrix} $$


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.