Let $V$ be a finite dimensional vector space and let $T:V\rightarrow V$ be a linear map. Suppose that $T$ is diagonalizable. Show that $Ch_T(T)=0$ without using the Cayley-Hamilton theorem. We are told as a hint to use: $f(T)\vec{v}=f(\lambda)\vec{v}$ where $\lambda\text{ and }\vec{v}$ are an eigenvalue and corresponding eigenvector of T. This was proved in the prior question. I would like to know if this is a valid proof.


Since $T$ is diagonalizable, $\exists ~~S\in\mathbb{R}^{n\times n}\text{ such that }D=S^{-1}TS$ Where $D$ is a diagonal matrix. Thus Matrices $D$ and $T$ are similar and they have the same characteristic equation and the same eigenvalues i.e. $Ch_T=Ch_D\text{ and } \lambda_T=\lambda_D$.

We have that $f(T)\vec{v}=f(\lambda)\vec{v}$, now let $f = Ch_T $.


$$Ch_T(T)\vec{v}=Ch_D(\lambda_D)\vec{v}$$ Since $D$ is diagonal $Ch_D(\lambda_D)=0$ because there will be at least one multiple of zero in its factors. Thus:$$Ch_T(T)\vec{v}=0$$And since $\vec{v}$ is non-zero (its an eigenvector) this implies that $Ch_T(T)=0._{~~\square}$

Attempt 2:

Since $T$ is diagonalizable, there exists a basis $B$ consisting of eigenvectors of $T$ such that $[T]_B$ is diagonal. Let $B=\{b_1,b_2,\dots,b_n\}$. In the basis $B$ the characteristic equation of $T$ has the following factorisation: $Ch_T(x)=(x-\lambda_1)(x-\lambda_2)\dots(x-\lambda_n)$. Need to show that $Ch_T(T)\vec{v}=0$. Also note that $v=\sum_{k=1}^n a_kb_k$ where $a_k$ is an element of the field.

$$Ch_T(T)\vec{v}=\prod_{i=1}^n[T-\lambda_iI]\sum_{k=1}^n[ a_kb_k]$$ $$~~~~~~~~~~~~~~~~~~=\sum_{k=1}^n\prod_{i=1}^n[Ta_kb_k-\lambda_iIa_kb_k]$$ $$~~~~~~~~~~~~~~~~~~=\sum_{k=1}^n\prod_{i=1}^n[a_k\lambda_kb_k-a_k\lambda_ib_k]$$ $$~~~~~~~~~~~~=\sum_{k=1}^n\prod_{i=1}^n(\lambda_k-\lambda_i)a_kb_k$$ $~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~=0_{~~\square}$

That last step is because in every term in the sum, there will be a factor of zero in the product when k=i.

  • $\begingroup$ You want be clear to distinguish between a linear map and a matrix. For example, it doesn't make sense to diagonalize a linear map $T$, you would instead be finding a basis under which the linear map $T$ is represented by a diagonal matrix. You are switching back and forth between the two points of view here, which is mathematically sort of awkward. $\endgroup$ – Morgan Rodgers Nov 19 '15 at 13:00
  • $\begingroup$ I thought diagonalizing a linear map and finding a basis in which it is a diagonal matrix were the same thing :S $\endgroup$ – NormalsNotFar Nov 19 '15 at 13:06
  • $\begingroup$ Not exactly; they involve the same concept, but a linear map and a matrix are different types of objects. You cannot multiply a linear map by a matrix to get a matrix, for example. $\endgroup$ – Morgan Rodgers Nov 19 '15 at 13:08

Your approach will lead to the right answer (but you may need to add the facts about an eigenbasis mentioned by @MorganRodgers)

An alternate approach is to observe that $A^n=SD^nS^{-1}$ where $A$ is the matrix of the linear transformation $T$ (after choosing a basis). Then, plugging in $A$ into the characteristic polynomial, you have $$ Ch_T(A)=S\begin{bmatrix}Ch_T(\lambda_1)\\&\ddots\\&&Ch_T(\lambda_n)\end{bmatrix}S^{-1} $$ But, since the roots of the characteristic polynomial are the eigenvalues, the diagonal matrix is actually the zero matrix. Therefore, $Ch_T(A)=0$.

| cite | improve this answer | |

Let $P(X)$ be the characteristic polynomial of an $n \times n$ matrix $A$. You want to show that $$P(A)v = 0 \mbox{ for all } v \in \mathbb{R}^{n}.$$ To do this, use the fact that $A$ is diagonalizable (that is, that there exists a basis for $V$ consisting entirely of eigenvectors of $A$) along with the factorization $$P(X) = (x-\lambda_{1})(x-\lambda_{2})\cdots(x-\lambda_{n})$$ where the $\lambda_{i}$ are the eigenvalues of $A$ (possibly with repeats).

| cite | improve this answer | |
  • $\begingroup$ Hi Morgan, I think this the approach I took in my attempt, could you check if it is valid? Thanks. $\endgroup$ – NormalsNotFar Nov 19 '15 at 13:01
  • $\begingroup$ @Pondzosaurusrex I don't find it especially clear, I don't understand what $f$ is or what it is used for. Nor is it clear where $v$ comes from. You really need to show that $P(A)v = 0$ for every possible choice of $v \in V$ to guarantee that $P(A) = 0$. $\endgroup$ – Morgan Rodgers Nov 19 '15 at 13:06
  • 1
    $\begingroup$ I agree, I think that one of the main steps that is missing in the original attempt (and is present in this answer) is that there is an eigenbasis, so that every $v\in V$ can be written in terms of eigenvectors. $\endgroup$ – Michael Burr Nov 19 '15 at 13:08
  • $\begingroup$ Ok I will make another attempt, using the advice you both have given me. I will update the original post. If you would like a bit more background knowledge of what $f$ is and $v$ see this link: math.stackexchange.com/questions/1535883/… $\endgroup$ – NormalsNotFar Nov 19 '15 at 13:23
  • $\begingroup$ Ok how does the second attempt look now? $\endgroup$ – NormalsNotFar Nov 19 '15 at 14:18

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.