Let $V$ be any $n$-dimensional vector space and $W = M_2(\mathbb{C})$.

(a) Construct all linear maps $T: V \to V$ such that $T^2 = I_n$.

For $T^2 = I_n$, taking $p(X) = X^2 - 1$ we must have $p(T) = 0$. Since the minimal polynomial of $T$ divides $p$ and contains all eigenvalues, $T$ must be diagonalizable with eigenvalues $1$ and $-1$. Since the minimal polynomial is of degree $\leq 2$, the largest Jordan block of the Jordan Form of $T$ must have order $2$. Then the maps satisfying $T^2 = I_n$ are the maps with eigenvalues $1$ and $-1$ such that the Jordan Normal Form is $J = (J^1,...,J^1)$ where each Jordan block $J^{i}$ have order $\leq 2$.

I'm not sure about my answer to this item.

(b) Find the Jordan Normal Form of $T: W \to W$ given by $T(X) = AX - XA$ with $$A = \left(\begin{array}{cc} 0 & 1 \\ 0 & 0 \end{array}\right).$$

We have to find the characteristical polynomial of $T$. If

$$X = \left(\begin{array}{cc} x & y \\ z & w \end{array}\right),$$


$$T(X) = \left(\begin{array}{cc} z & w-x \\ 0 & -z \end{array}\right)$$

I have problem to find the matrix associate to $T$ so, I tried to use $W \simeq \mathbb{R}^4$ and define $T: \mathbb{R}^4 \to \mathbb{R}^4$ by

$$T(x,y,z,w) = (z, w-z,0,-z).$$

Thus, $$T(1,0,0,0) = (0,0,0,0) = 0e_1 + 0e_2 + 0e_3 + 0e_4,$$ $$T(0,1,0,0) = (0,0,0,0) = 0e_1 + 0e_2 + 0e_3 + 0e_4,$$ $$T(0,0,1,0) = (1,-1,0,-1) = 1e_1 -1e_2 + 0e_3 -1e_4,$$ $$T(0,0,0,1) = (0,1,0,0) = 0e_1 + 1e_2 + 0e_3 + 0e_4.$$

So, the matrix is $$T = \left(\begin{array}{cccc} 0 & 0 & 1 & 0 \\ 0 & 0 & -1 & 1 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & -1 & 0 \end{array}\right).$$

Is that matrix correct? If yes, I can continue. If no, I would like some help.

  • 1
    $\begingroup$ $p=(X-1)(X+1)$. Therefore T is diagonalizable and there are no Jordan block of size 2. $\endgroup$ – TomTom314 Sep 25 at 14:13
  • 1
    $\begingroup$ By the way, $W\cong \Bbb C^4$, not $\Bbb R^4$. So you should have $T:\Bbb C^4\to \Bbb C^4$. $\endgroup$ – WE Tutorial School Sep 25 at 15:49

For (a), TomTom314 already mentioned that $T$ is diagonalizable with eigenvalues $\pm1$. In the case that all eigenvalues are the same, it is clear that $T=\pm I$. But for the rest, I think the best you can say about $T$ is $T=XDX^{-1}$ where $X$ is invertible and $D$ is a diagonal matrix of the form $$\operatorname{diag}(-1,-1,\ldots,-1,1,1,\ldots,1)$$ where $-1$ appears $p$ times and $1$ appears $q$ times ($p,q$ are positive integers such that $p+q=n$).

However, if this is what the problem means, we can find a 1-1 correspondence between all such $T$ and all projections on $V$. For a projection $P:V\to V$, set $T_P=2P-I$. Then $T_P^2=I$. For a map $T$ such that $T^2=I$, define $P_T=\frac{1}{2}(T+I)$ so $P_T^2=P_T$, making $P_T$ a projection.

For (b), the matrix of $T$ in your chosen basis is correct, but I'd like to point out that $T$ is nilpotent. So all eigenvalues are $0$. Note that $$T^2X=A(AX-XA)-(AX-XA)A=A^2X-2AXA-XA^2$$ and since $A^2=O$, we have $$T^2X=-2AXA.$$ That is, $$T^3X=A(-2AXA)-(-2AXA)A=-2A^2XA+2AXA^2=O.$$ Since $T^2A^t=-2AA^tA\ne O$, $T$ is nilpotent of depth $3$. That is $T$ has a Jordan block of size $3$, and since the space $M_2(\Bbb C)$ is $4$-dimensional, $T$ has another Jordan block of size $1$. That is, a Jordan normal form of $T$ is $$\left(\begin{array}{ccc|c}0&1&0&0\\0&0&1&0\\0&0&0&0\\\hline0&0&0&0\end{array}\right).$$



(1) As you've written, $\require{cancel}T^2 = \operatorname{id}$ implies that the minimal polynomial $m(x)$ of $T$ divides $x^2 - 1 = (x - 1) (x + 1)$. But as pointed out in the comments, (at least provided that the field $\Bbb F$ underlying $V$ does not have characteristic $2$) the linear factors of $m(x)$ do not repeat, so $T$ must be diagonalizable.

Thus, there is a basis of $V$ with respect to which the matrix representation $[T]$ of $T$ has the form $$\pmatrix{I_k \\ & -I_{n - k}} .$$ Put another way, if we fix any basis of $V$, the matrix representations $[T]$ of the linear transformations $T$ satisfying $T^2 = \operatorname{id}$ are precisely those similar to one of the above three matrices, that is, $\pm I$ and those of the form $$P^{-1} \pmatrix{I_k \\ & -I_{n - k}} P .$$

Remark The restriction on the characteristic is necessary: In characteristic $2$, $x^2 - 1 = (x - 1)^2$, so for example, that the nondiagonalizable matrix $A = \pmatrix{1&1\\&1}$ satisfies $A^2 = I$.

(2) In this case we can compute the Jordan Normal Form without working directly with the entries of $X$--though if you're not familiar with Lie algebras (which is surely the most important context wherein transformations of this form appear), computing $T^k(X)$ in terms of entries helps motivate this approach:

Notice that $A^2 = 0$---this makes easier the computation of powers of $T$: We have \begin{multline}T^2(X) = T(AX - XA) = A(AX - XA) - (AX - XA)A \\= \cancel{A^2 X} - 2 A X A + \cancel{X A^2} = -2 A X A,\end{multline} and computing similarly gives $$T^3 = 0 .$$

Since $T$ is thus nilpotent, its only eigenvalue is $0$, and all that remains to find is the size of the Jordan blocks. But since $T^3 = 0$ and $T^2 \neq 0$, the Jordan normal form contains a $3 \times 3$ block, $J_3(0)$, so the remaining block is $J_1(0) = \{0\}$. Thus, the Jordan normal form of $T$ is $$J_3(0) \oplus J_1(0) = \pmatrix{0&1&\cdot\\\cdot&0&1\\ \cdot&\cdot&0\\&&&0} .$$

Remark Our computations of $T^2, T^3$ used only the definition of multiplication and the fact that $A^2 = 0$ (and not the specific form of $A$). But $A$ is the Jordan form of any nonzero $2 \times 2$ nilpotent matrix, so our conclusion applies to any such matrix.

You might find it instructive to repeat the exercise replacing $A$ with a general $2 \times 2$ matrix. The observation implicit in the previous remark shows that it's enough to consider just the $A$ in Jordan form, so those of the forms $$\lambda I, \quad \pmatrix{\lambda&1\\&\lambda}, \quad \operatorname{diag}(\lambda, \mu) .$$


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.