MathOverflow is a question and answer site for professional mathematicians. 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

There is a vast literature on Jacobi matrices, I just don't know where to start looking. I'm interested in estimating the largest eigenvalue of the $n\times n$ periodic Jacobi matrix $D+P+P^{-1}$, where $P$ is the matrix of the cyclic permutation of coordinates, and $D$ is the diagonal matrix with diagonal entries $2\cos\frac{2k\pi}{n}$ (where $0\leq k\leq n-1$). I can't help thinking that this has been already considered...

share|cite|improve this question
up vote 1 down vote accepted

It's roughly $2 - O(1/n^2)$. For an appropriate choice of $K$ evaluate $(D + P + P^{-1}) u$ for $$ u(k) = \begin{cases} 1 - \frac{|k|}{K} & |k| \leq K \\\ 0 & otherwise.\end{cases} $$ An application of the uncertainty principle shows that this is optimal up to constants.

The main point why this works is that $2 \cos( 2 \pi \frac{k}{n})$ is almost constant in $k$. If you would instead take for some $\ell$ coprime to $n$ the potential $2 \cos(2 \pi \frac{\ell k}{n})$. The question is much harder.

share|cite|improve this answer
Don't you mean $4-O(1/n^2)$ ? By the way, let $A_\ell$ be the matrix where, as you suggest, you replace the potential $2\cos\frac{2\pi k}{n}$ by $2\cos\frac{2\pi k\ell}{n}$. Is it true that the maximum of the $\|A_\ell\|$'s is attained for $\ell=1$? – Alain Valette May 21 '11 at 19:37
It's 4 not 2. The maximum is at l = 1. See l is roughly the vertical axis in the Hofstadter butterfly. – Helge May 21 '11 at 20:13
Gosh! And I knew it 15 years ago, as I needed it in: On the spectrum of a random walk on the discrete Heisenberg group and the norm of Harper's operator Authors: Beguin C.1; Valette A.; Zuk A. Source: Journal of Geometry and Physics, Volume 21, Number 4, March 1997 , pp. 337-356(20) Thanks anyway for your help! – Alain Valette May 21 '11 at 20:47

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.