Fix an integer $d$, let $\mathbb{U}_d$ be the $d\times d$ unitary group.

For any $U\in \mathbb{U}_d$, define $\Omega(U)$ be the length of the smallest arc containing all the eigenvalues of $U$ on the unit circle.

Now, we are interested in the probability density function of $\Omega(U)$ as $U$ being distributed as Haar measure.

What is the probability that $\Omega(U)$ is smaller than some given constant $\epsilon$?


This amounts to the question what is the probability distribution of the largest gap $\Delta=2\pi-\Omega$ between the eigenphases in the circular unitary ensemble. It was addressed in Extreme gaps between eigenvalues of random matrices (2010). Asymptotically for large $d$, the distribution $P(\Delta)$ is peaked around $\Delta_0=d^{-1}\sqrt{32\log d}$, with Poissonian fluctuations around this value.

| cite | improve this answer | |
  • $\begingroup$ Thanks! Is there explicit probability density function? $\endgroup$ – gondolf May 3 '16 at 13:51
  • $\begingroup$ it's like all spacing distributions in random matrix theory, you only get explicit results asymptotically for large dimensionality. $\endgroup$ – Carlo Beenakker May 3 '16 at 14:35
  • $\begingroup$ Thank you! Just would like to make sure that the " Extreme gaps" is indeed "the length of the smallest arc containing all the eigenvalues”? $\endgroup$ – gondolf May 3 '16 at 17:30
  • $\begingroup$ "extreme gap" is "length of largest arc containing no eigenvalues", so that's the $2\pi$ minus the length of the smallest arc containing all eigenvalues, isn't it? $\endgroup$ – Carlo Beenakker May 3 '16 at 19:37
  • $\begingroup$ Very promising! $\endgroup$ – gondolf May 3 '16 at 21:03

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.