We are given vectors $a_1, ... , a_n$, which are linearly independent and a set of equations about a linear map $A: \mathbb{R}^n \rightarrow \mathbb{R}^n$;

$Aa_1 = a_2$, $A^2a_1 = a_3$, ... , $A^na_1 = a_1$

and are supposed to find its eigenvalues, eigenvectors and the characteristic polynomial from the given information.

We can see that the set of equations can also be written as:

$Aa_1 = a_2$, $Aa_2 = a_3$, ... , $Aa_n = a_1$

and by summing all of them we find the first eigenvalue to be 1 (and the corresponding eigenvector $a_1 + a_2 + ... + a_n$).

I'm now having a problem finding the other eigenvalues. I wrote down $A$ for a general $x$, which can be written as $x = b_1a_1 + b_2a_2 + ... + b_na_n$, where $b_1, ..., b_n \in \mathbb{R}$, since $a_1, ... , a_n$ are linearly independent in a $n$-dimensional space.

From that I got:

$Ax = A(b_1a_1 + b_2a_2 + ... + b_na_n) = b_1Aa_1 + ... + b_nAa_n = b_1a_2 + ... + b_{n-1}a_n + b_na_1$

so for x to be an eigenvector it has to be

$b_na_1 + b_1a_2 + ... + b_{n-1}a_n = \lambda (b_1a_1 + b_2a_2 + ... + b_na_n)$

which gives another system of equations to find $\lambda$, the eigenvalues. It seems all eigenvalues will be $1$ or $-1$ (I'm just assuming, based on the equations I got), I just have trouble proving it.

Thank you for any help with my problem. :)


Actually,you can denote $A$ as a special circulant matrix as $J$,e.g. \begin{pmatrix} 0 & 1 & 0 &\cdots & 0\\ 0 & 0 & 1 &\cdots & 0\\ \vdots & \vdots & \vdots &\ddots & \vdots\\ 0 & 0 & 0 & \cdots & 1 \\ 1 & 0 & 0 & \cdots & 0 \end{pmatrix} The eigenvalues of $J$ is obvious. The characteristic polynomial would be $f(\lambda)=\lambda^n+(-1)^n$

Hope that my short answer may help.

Correction: $A^T$ $=J$,but it doesn't matter because $A \sim A^T$.($\sim$ means similar).


First let me answer your question over $\mathbb{C}$.

You have that $A^n=I$ and $A^{n-1}\neq I$ so the minimal and the characteristic polynomials are equal to $X^n-1$.

The eigenvalues are therefore $\zeta^k$ for $k=0,\dots, n-1$, where $\zeta:=\exp\frac{2\pi i}{n}$.

The eigenvector corresponding to $\zeta^k$ is $\sum_{l=1}^{n} \zeta^{k(l-1)}a_l$; this is easily checked.

Now over $\mathbb{R}$ there are only one or two eigenvalues: always $+1$, and $-1$ when $n$ is even. The corresponding eigenvectors are unchanged.


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.