1: \begin{abstract}

2: We combine concepts from random matrix theory and free probability

3: together with ideas from the theory of commutator length in groups

4: and maps from surfaces, and establish new connections between the

5: two. 

6: 

7: More particularly, we study measures induced by free words on the

8: unitary groups $\U\left(n\right)$. Every word $w$ in the free group

9: $\F_{r}$ on $r$ generators determines a word map from $\U\left(n\right)^{r}$

10: to $\U\left(n\right)$, defined by substitutions. The $w$-measure

11: on $\U\left(n\right)$ is defined as the pushforward via this word

12: map of the Haar measure on $\U\left(n\right)^{r}$.

13: 

14: Let $\trw\left(n\right)$ denote the expected trace of a random unitary

15: matrix sampled from $\U\left(n\right)$ according to the $w$-measure.

16: It was shown by Voiculescu \cite{VOIC} that for $w\ne1$ this expected

17: trace is $o\left(n\right)$ asymptotically in $n$. We relate the

18: numbers $\trw\left(n\right)$ to the theory of commutator length of

19: words and obtain a much stronger statement: $\trw\left(n\right)=O\left(n^{1-2g}\right)$,

20: where $g$ is the commutator length of $w$. Moreover, we analyze

21: the number\linebreak{}

22: $\lim_{n\to\infty}n^{2g-1}\cdot\trw\left(n\right)$ and show it is

23: an integer which, roughly, counts the number of (equivalence classes

24: of) solutions to the equation $\left[u_{1},v_{1}\right]\ldots\left[u_{g},v_{g}\right]=w$

25: with $u_{i},v_{i}\in\F_{r}$. 

26: 

27: Similar results are obtained for finite sets of words and their

28: commutator length, and we deduce that one can {}``hear'' the stable

29: commutator length of a word by {}``listening'' to its unitary measures.

30: \end{abstract}