Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Reading through this paper I've come across a statement that I don't follow, could someone give some pointers/hints?

Let $A$ be the $2n\times 2n$ matrix given by

$$A=(I_n\otimes F)+(G\otimes H),\quad\quad(*)$$


$$F=\begin{bmatrix}0&1\\-c_2c_4&-(c_2+c_4)\end{bmatrix},\quad\quad H=\begin{bmatrix}0&0\\ c_1c_3\alpha&0\end{bmatrix},$$

with $\alpha,c_1,\dots,c_4\in\mathbb{R},$ and $G$ is the $n\times n$ circulant matrix


The authors write: "Fourier-diagonalizing the circulant matrix $G$, we obtain the following complex quadratic equation for the eigenvalues $\lambda_l$ of the matrix $A$:

$$\lambda_l^2+(c_2+c_4)\lambda_l+c_2c_4\alpha\left(1-\frac{c_1c_3}{c_2c_4}\alpha e^{i\frac{2\pi}{n}l}\right),\quad\quad (**)$$

with $l=0,\dots,n-1$."

From the above I assume they showed that the characteristic polynomial of $A$ can be factored into $n$ quadratic polynomials in $(**)$. Fourier-diagonalizing $G$ gives

$$G=\frac{1}{n}U^*\text{diag}(u)U,\quad\quad U=\begin{bmatrix}\omega_n^{0\cdot0}&\omega_n^{0\cdot1}&\dots&\omega_n^{0\cdot(n-1)}\\ \omega_n^{1\cdot0}&\omega_n^{1\cdot1}&\dots&\omega_n^{1\cdot(n-1)}\\ \vdots&\vdots&\ddots&\vdots\\ \omega_n^{(n-1)\cdot0}&\omega_n^{(n-1)\cdot1}&\dots&\omega_n^{(n-1)\cdot(n-1)}\end{bmatrix},\quad\quad u =\begin{bmatrix}\omega_n^{0\cdot1}\\\omega_n^{1\cdot1}\\\vdots\\\omega_n^{(n-1)\cdot1}\end{bmatrix},$$

where $\omega_n=e^{-\frac{2\pi}{n}}$, and diag$(u)$ denotes the diagonal matrix with vector $u$ on its diagonal. This is as far as I get, any hints on how to proceed?

Thanks in advance.

share|cite|improve this question

Clearly, the formula is false. Indeed, let $\alpha=0$ and assume $c_2c_4\not=0$.Then $A=I\otimes F$ is invertible and necessarily $\lambda_l\not= 0$, that is a contradiction. On the other hand, your formula concerning $\omega_n$ is false. Indeed $G^n=I$, then $w_n=exp(2i\pi/n)$. Here $U$ has nothing to do because $I$ is invariant by a change of basis. More precisely you can take $U=I$. The calculation is easy; we find (we note $\omega$ for $\omega_n$) a diagonal of $2\times 2$ matrices in the form $\begin{pmatrix} 0&1\\-c_2c_4+\omega^lc_1c_3\alpha&-(c_2+c_4)\end{pmatrix}$. Finally the eigenvalues are the roots of the polynomials $\lambda_l^2+(c_2+c_4)\lambda_l+c_2c_4-\omega^lc_1c_3\alpha$ and $\alpha$ is not in the factor.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.