I need to find the eigenvalues and eigenvectors of $A$ in $\mathbb{R}^3$ such that $Ax = |a|^2x-(a \cdot x)a$ where $a$ is a given constant vector.

I found $A-\lambda I$ in the usual $\mathbb{R}^3$ basis:

\begin{bmatrix} |a|^2-a_x^2-\lambda & -a_xa_y & -a_xa_z \\ -a_xa_y & |a|^2-a_y^2-\lambda & -a_ya_z \\ -a_xa_z & -a_ya_z & |a|^2-a_z^2-\lambda \end{bmatrix}

The determinant of this matrix is $(\lambda)(\lambda-x^2-y^2-z^2)^2$ (Wolfram). This gives me eigenvalues of $0$ and $|a|^2$. Is this correct? If so, how do I find the eigenvectors in this system? Typically for the $\lambda=0$ case I would simply reduce A to row-echelon form and find the vector. Setting $\lambda=0$ gives me:

\begin{bmatrix} a_y^2 + a_z^2 & -a_xa_y & -a_xa_z \\ -a_xa_y & a_x^2 + a_z^2 & -a_ya_z \\ -a_xa_z & -a_ya_z & a_x^2 + a_y^2 \end{bmatrix}

I don't see how to reduce this matrix or proceed from here to find eigenvectors.


I think it's easier not to use matrix representation.

Suppose $x$ is an eigenvector. Then $Ax = \lambda x$, i.e.,

$$|a|^2 x - (a.x)a = \lambda x \ \ \hbox{ or } \ \ (|a|^2 - \lambda)x = (a.x) a$$

If $x$ is parallel with $a$, then $x = a$ is an eigenvector and

$$(|a|^2 - \lambda)a = |a|^2a \ \ \Longrightarrow \ \ \lambda = 0.$$

That is, $a$ is a basis vector for the kernel of $A$. Inspecting $A$, it is the only basis vector.

If $x$ and $a$ are not parallel, it must be that $a.x = 0$ with eigenvalue $\lambda = a^2$. I'll leave it to you to find the corresponding eigenvectors/eigenspace.


We can write $$ \mathsf{A}\mathbf{\cdot x}=\mathbf{|a|}^{2}\{\mathsf{U}-\mathbf{e}_{3} \mathbf{e}_{3}\}\cdot \mathbf{x} $$ where $\mathsf{U}$ is the unit $3\times 3$-matrix and $\mathbf{e}_{3}$ the unit vector along $\mathbf{a}$. With $\{\mathbf{e}_{1},\mathbf{e}_{2}, \mathbf{e}_{3}\}$ an orthonormal basis, we have $$ \mathsf{U}=\sum_{j=1}^{3}\mathbf{e}_{j}\mathbf{e}_{j}=\mathsf{U}_{1,2}+ \mathbf{e}_{3}\mathbf{e}_{3} $$ so $\mathsf{A}$ decomposes in the direct sum of $\mathbf{|a|}^{2}$ times the unit operator on the subspace spanned by $\mathbf{e}_{1}$ and $\mathbf{e}_{2} $ and the zero operator on the subspace spanned by $\mathbf{e}_{3}$. Thus there are two independent eigenvectors $\mathbf{e}_{1}$ and $\mathbf{e}_{2} $ (or appropriate linear combinations) at the eigenvalue $\mathbf{|a|}^{2}$.

By the way, another form of $\mathsf{A}\mathbf{\cdot x}$ is $\mathbf{a}% \times (\mathbf{a\times x)}$.


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.