2
$\begingroup$

I came across this question:

Consider the plane as the real linear space $R^2$, and let $T$ be a rotation of $R^2$ through an angle of $\pi/2$ radians about the origin. Although $T$ has no eigenvectors, prove that every nonzero vector is an eigenvector for $T^2$ for $T^2$.

1). My main question is how to do you show that $T$ has no eigenvectors?

2). Second question: does this suffice as a proof of the above statement.

Let $T^2 = \left(\begin{array}{cc}-1 & 0 \\0 & -1\end{array}\right)$. You get this by matrix multiplying the rotation matrix for $R^2$ twice. Let $a,b \in$ Reals, and $x = (a,b)^T$. $T^2[x] = -1x$. Thus, for every non-zero vector $x$, it is an eigenvector (with eigenvalue of -1).

$\endgroup$
1
  • $\begingroup$ You have to understand the geometric meaning of rotation and eigenvector... A rotation doesn't respect any line (that is it never sends any line to itself). An eigenvector means that the operation does exactly that; send the line that contains the eigenvector to itself... $\endgroup$ Apr 10, 2015 at 1:56

3 Answers 3

2
$\begingroup$

Your second proof should suffice, although I don't know what $(a, b)$ and $\mathbf{x}$ are both doing in there -- are $a, b$ the elements of the vector $\mathbf{x}$? You could just note that $T^2 = -I$, hence $T^2\mathbf{x} = -I\mathbf{x} = -\mathbf{x}$.

You can show that $T$ has no eigenvectors by showing that it has no (real) eigenvalues. Consider the eigenvalue equation $$\det{(T - \lambda I)} = 0$$

Since $$T = \begin{bmatrix}0 & -1\\1 & 0\end{bmatrix}$$

we have $$T - \lambda I = \begin{bmatrix}0 & -1\\1 & 0\end{bmatrix} - \begin{bmatrix}\lambda & 0\\0 & \lambda\end{bmatrix} = \begin{bmatrix}-\lambda & -1\\1 & -\lambda\end{bmatrix}$$

we can see that $\det({T - \lambda I}) = \lambda^2 + 1$. The solutions of this polynomial in $\lambda$ are the eigenvalues of $T$. What are the (real) solutions? None. Clearly the parabola $\lambda^2 + 1$ never touches the $x$-axis!

$\endgroup$
1
$\begingroup$

For (1), you might either write down a formula for $T(x, y)$ and solve $T(x, y) = \lambda(x, y)$ "manually", or find the standard matrix of $T$ and compute the characteristic polynomial.

Your answer for (2) is correct.

$\endgroup$
1
$\begingroup$

the rotation matrix $A = \pmatrix{0&-1\\1&0}$ has no real eigenvalues and no eigenvectors with all real components; however, it has complex eigenvalues $\pm i$ and the corresponding eigenvectors are $\pmatrix{\pm i\\1}.$

one reason that $A$ has no real eigenvalue is $Au$ is never parallel to $u$ for any $u \neq 0$ because $A$ rotates $u$ by $90^\circ$ counter clockwise.

$\endgroup$

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.