3

Consider the following matrix $A=\pmatrix{0 &1\\ -1&\frac{1}{3}}$ and put $e_1:=(1,0)^t$. How can I find the elements of the orbit $A^ne_1$, for $n\in\mathbb N$?

It seems to me that $A$ has infinite order in $SL_2(\mathbb R)$ and moreover I don't see any closed formula for the powers of $A$.

manifold
  • 1,843

1 Answers1

4

First let's note that characteristic polynomial of $A$ is $P_{A}(t) = t^{2}-\frac{1}{3}t+1$, which is irreducible over $\mathbb{R}[t]$ since it's root are $\frac{1}{6}(1 \pm i\sqrt35)$; In particular, it's irreducible over $\mathbb{Q}[t]$.

Since the minimal polynomial of $A$,denoted with $m_{A}$ divide the characteristic polynomial but the second one it's irreducible,

We can now assert that the minimal polynomial of $A$ and the characteristic polynomial coincide.

If $A$ would have finite order,for the Hamilton-Cayley theorem it would follow that $m_{A}(t) \mid t^{n}-1$ for some $n \in \mathbb{N}-\{0\}$

We can notice that $t^{n}-1 = \prod_\limits{d \mid n} \phi_{d}(t)$

(Proving that cyclotomic polynomials have integer coefficients)

Where $\phi_{d}(t)$ are the $d-th$ cyclotomic polynomial, and this factorization lie in $\mathbb{Z}[t]$.

Since we noticed that $P_{A}(t)$ is irreducibile over $\mathbb{Q}[t]$ which is (UFD),

If $P_{A}(t)$ would divide $t^{n}-1$ it would coincide with one of the irreducible factors of $t^{n}-1$, but those lie in $\mathbb{Z}[t]$, while $P_{A}(t)$ don't.

Edit : If orb($e_{1}$) would be finite, even the orbit of $x=Ae_{1}$ would be finite.

But $e_{1},x$ are linearly independent and we have shown that $A^{n}e_{1} = e_{1}$ and $A^{m}x=x$ for some $m,n$. It would follow that $A^{nm} = I_{d}$

Ps : If you're interested in finding the powers of $A$, a way to start could be noticing that the Jordan form of $A$ is indeed, a rotation, which powers are relatively easily computable. (Questions of this kind or similar have been answered on MSE Powers of Rotation matrix)

jacopoburelli
  • 5,564
  • 3
  • 12
  • 32
  • 1
    $A$ can have infinite order, but $e_1$ finite orbit... – manifold Apr 10 '20 at 20:36
  • 1
    @manifold I see, I din't really specified the conclusion. I edited the answer,now should work. – jacopoburelli Apr 10 '20 at 21:04
  • 1
    It is interesting to note 2 things. The first one is that your proof works for every matrix of the form $\left(\begin{matrix}0&1\-1& \frac{1}{n} \end{matrix}\right)$ with $n\in \mathbb Z-{0}$. The second one is that you also show that every vector has an infinite orbit under the action of this type of matrices – Menezio Apr 11 '20 at 19:25