4

Be $M \in M_2(\mathbb{R})$ with $\det(M) = 1$

If $|\text{tr}(M)| < 2$, Prove that exist $P \in M_2(\mathbb{R})$ with $\det(P)=1$ and exist $\alpha \in \mathbb{R}$ such that:

$ M = P\cdot \begin{bmatrix} \cos(\alpha) & -\sin(\alpha) \\ \sin(\alpha) & \cos(\alpha) \end{bmatrix} \cdot P^{-1}$

$\\$

My attempt to solution:

I could not solve the problem. And I do not know how to continue.

$1)$ I started analising the caracteristic polynomial of M:

$P(x)_M = x^2-\text{tr}(M)\cdot x + \det(M)$

And we know that $\det(M) = 1$ and $|\text{tr}(M)| < 2$, implies that $\Delta = (\text{tr}(M))^2-4 \det(M)<0$

So, both eigenvalues of $M$ are complexes and distinct, and one is conjugate of another.

$2)$ This matrix remember me the rotation matrix, but i don't know how to use that fact.

I do not know what to do. I tried for a long time, but I really could not resolve.

terran
  • 1,380
HZLJ
  • 141
  • 5
  • Parameterizing the eigenvalues in polar form as $r_1e^{i\theta}$ and $r_2e^{-i\theta}$, the fact that the determinant equals 1 means that $r_2=1/r_1$. Two distinct eigenvalues means that the matrix is diagonalizable. Start in the basis of eigenvectors (where the matrix is diagonal), and see if you can construct a $P$ that turns this into the rotation matrix above. Note that $\theta$ will be equal to $\alpha$. I believe you can show that $r_2=r_1=1$, but I haven't quite figured that out yet. The real problem is then making a corresponding $P$ that is real... – march Oct 22 '21 at 16:31

1 Answers1

1

Denote by $\lambda_1$ and $\lambda_2 = \overline{\lambda_1}$ the two complex eigenvalues. Since $$\det(M) = \lambda_1 \lambda_2 = |\lambda_1|^2 = 1$$ this means that we can write $\lambda_1 = e^{i\alpha}$ and $\lambda_2 = e^{-i\alpha}$ for some $\alpha \in \mathbb{R}$. Since the eigenvalues are distinct, this means that $M$ is diagonalizable over $\mathbb{C}$. Let $v_1 \in \mathbb{C}^2$ be a complex eigenvector corresponding to $\lambda_1$. Then since $M$ is real we have $$ M \overline{v_1} = \overline{Mv_1} = \overline{\lambda_1 v_1} = \overline{\lambda_1} \overline{v_1} = \lambda_2 \overline{v_1} $$ so $\overline{v_1}$ is an eigenvector of $M$ corresponding to $\lambda_2$. Set $$ u_1 := \frac{v_1 - \overline{v_1}}{2}, u_2 := \frac{v_1 + \overline{v_1}}{2i}. $$ That is, $u_1$ and $u_2$ are the real and imaginary parts of the eigenvector $v_1 = u_1 + iu_2$. Note that $u_1,u_2$ must be linearly independent over $\mathbb{R}$. Why? If $au_1 + bu_2 = 0$ for $a,b \in \mathbb{R}$ then $$ au_1 + bu_2 = a \cdot \frac{v_1 - \overline{v_1}}{2} + b \cdot \frac{v_1 + \overline{v_1}}{2i} = \frac{1}{2} \left( (a - ib) v_1 -(a + ib) \overline{v_1} \right) = 0. $$ Since $v_1$ and $\overline{v_1}$ are linearly independent, we must have $$ a - ib = a + ib = 0 \implies a = b = 0. $$

Let's see how $M$ acts on $u_1$:

$$ Mu_1 = \frac{\lambda_1 v_1 - \overline{\lambda_1} \overline{v_1}}{2} = \frac{\lambda_1 (u_1 + iu_2) - \overline{\lambda_1}(u_1 - iu_2)}{2} \\ = \frac{\lambda_1 - \overline{\lambda_1}}{2}u_1 + \frac{\lambda_1 + \overline{\lambda_1}}{2i} u_2 = \cos \alpha \cdot u_1 + \sin \alpha \cdot u_2. $$

A similar calculation shows that $$ Mu_2 = -\sin \alpha \cdot u_1 + \cos \alpha \cdot u_2. $$

This means that if you take $P \in M_2(\mathbb{R})$ whose columns are $u_1,u_2$ then $P$ will be invertible and $$P^{-1} M P = \begin{pmatrix} \cos \alpha & -\sin \alpha \\ \sin \alpha & \cos \alpha \end{pmatrix}. $$

levap
  • 65,634
  • 5
  • 79
  • 122