Proving that all $M_{2\times 2}(\mathbb{R})$ whose c.p does not split are similar to a positve multiple of a rotation matrix.

89 Views Asked by At

I want to prove that every matrix in $M_{2\times 2}(\mathbb{R})$ whose characteristic polynomial does not split is similar to a positive multiple of a rotation matrix, $$ s\begin{pmatrix} \cos \theta & -\sin \theta \\ \sin \theta & \cos \theta \end{pmatrix}. $$

I know that two matrices $A$ and $B$ are similar if there's an (invertible) matrix $Q$ such that $B = Q^{-1}AQ$. I think my issue is that I don't know how to operationalize the fact that the polynomial does not split, and furthermore how this is connected to the rotation matrix.

Any tips on how to proceed?

1

There are 1 best solutions below

0
On BEST ANSWER

One proof of this is as follows. First, note that every matrix of the form $$ M = \pmatrix{a&-b\\b&a} $$ (where $a,b$ are not both zero) is a positive multiple of some rotation matrix. Now, the characteristic polynomial of $A$ can be written in the form $$ p(x) = (x-a)^2 + b^2 $$ for some constants $a,b$ with $b$ non-zero. Let $B$ denote the matrix $B = A - aI$. The characteristic polynomial of $B$ must be of the form $q(x) = x^2 + b^2$.

Let $x \in \Bbb R^2$ be any non-zero vector. We note that $Bx = \lambda x$ cannot hold for any $\lambda \in \Bbb R$; otherwise, $(x-\lambda)$ would be factor of the characteristic polynomial $p(x)$. Thus, $x$ and $Bx$ are linearly independent. Consider the basis of $\Bbb R^2$ given by $$ \mathcal B = \{v_1,v_2\}, \qquad v_1 = x, \quad v_2 = \frac 1{b} Bx. $$ By the Cayley Hamilton theorem, we must have $q(B) = 0$, and hence $q(B)x = 0$, so that $$ (B^2 + b^2 I)x = 0 \implies B \left(\frac 1b x\right) = -bx. $$ Thus, we have $Bv_1 = bv_2$ and $Bv_2 = -bv_1$. Conclude that the matrix of $B$ relative to the basis $\mathcal B$ is given by $$ [B]_{\mathcal B} = \pmatrix{0&-b\\b&0}. $$ Conclude that the matrix of $A = aI + B$ is given by $$ [A]_{\mathcal B} = [aI + B]_{\mathcal B} = aI + [B]_{\mathcal B} = \pmatrix{a&-b\\b&a}. $$ Equivalently, if $Q$ is the matrix whose columns are $v_1,v_2$, then $$ Q^{-1}AQ = \pmatrix{a&-b\\b&a}, $$ which was what we wanted.