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


I am trying to understand the paper of Dembo-Vershik-Zeitouni, Large deviations for integer partitions. I am only interested in Theorem 2, which deals with the case of the uniform distribution. It describes a Large Deviation Principle LDP with a rate function $I:\widehat{DF}\to[0,\infty]$ of a form I expected to see, and speed $\sqrt n$. (Here, $\widehat{DF}$ is a space of functions that includes the ones describing the rims of partitions.) The LDP seems to imply that, for example, if I take the set $S_\varepsilon$ of functions $\varphi$ such that $I(\varphi)\geq \varepsilon$ then $$\mathbb P(S_\varepsilon)\approx \exp\left(-{\sqrt n}\varepsilon\right).$$

Since the limit shape is the only function for which $I$ vanishes, I expected the probability of $S_\varepsilon$ to decrease as $n\to\infty$, because ---as I had understood this limit shape business--- a greater proportion would be closer to the limit shape every time (and here we are keeping partitions far from the limit shape by requiring the value of $I$ to be greater than $\varepsilon$). However, in the asymptotics above, it seems that $\mathbb P(S_\varepsilon)\to 1$ as $n\to\infty$. What is going on?!

And this does happen, so there is no problem.

Thank you so much in advance!!

share|cite|improve this question
I edited your question to reflect your correction. You should delete your answer when you get a chance. Also, very interesting question. – Grant Rotskoff Apr 13 '12 at 15:19
Thanks. The thing is, my answer does answer my question. This question should be closed. – Zatrapilla Apr 13 '12 at 15:21

Your Answer


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

Browse other questions tagged or ask your own question.