Plancherel measure

Definition on the symmetric group

An important special case is the case of the finite symmetric group ${\displaystyle S_n}$, where ${\displaystyle n}$ is a positive integer. For this group, the set ${\displaystyle S_n^{\wedge }}$ of irreducible representations is in natural bijection with the set of integer partitions of ${\displaystyle n}$. For an irreducible representation associated with an integer partition ${\displaystyle \lambda }$, its dimension is known to be equal to ${\displaystyle f^{\lambda }}$, the number of standard Young tableaux of shape ${\displaystyle \lambda }$, so in this case Plancherel measure is often thought of as a measure on the set of integer partitions of given order n, given by

${\displaystyle \mu (\lambda )=\frac{(f^{\lambda }{)}^2}{n!}.}$

The fact that those probabilities sum up to 1 follows from the combinatorial identity

${\displaystyle \sum _{\lambda ⊢n}(f^{\lambda }{)}^2=n!,}$

which corresponds to the bijective nature of the Robinson–Schensted correspondence.

Application

Plancherel measure appears naturally in combinatorial and probabilistic problems, especially in the study of longest increasing subsequence of a random permutation ${\displaystyle \sigma }$. As a result of its importance in that area, in many current research papers the term Plancherel measure almost exclusively refers to the case of the symmetric group ${\displaystyle S_n}$.

Connection to longest increasing subsequence

Let ${\displaystyle L(\sigma )}$ denote the length of a longest increasing subsequence of a random permutation ${\displaystyle \sigma }$ in ${\displaystyle S_n}$ chosen according to the uniform distribution. Let ${\displaystyle \lambda }$ denote the shape of the corresponding Young tableaux related to ${\displaystyle \sigma }$ by the Robinson–Schensted correspondence. Then the following identity holds:

${\displaystyle L(\sigma )={\lambda }_1,}$

where ${\displaystyle {\lambda }_1}$ denotes the length of the first row of ${\displaystyle \lambda }$. Furthermore, from the fact that the Robinson–Schensted correspondence is bijective it follows that the distribution of ${\displaystyle \lambda }$ is exactly the Plancherel measure on ${\displaystyle S_n}$. So, to understand the behavior of ${\displaystyle L(\sigma )}$, it is natural to look at ${\displaystyle {\lambda }_1}$ with ${\displaystyle \lambda }$ chosen according to the Plancherel measure in ${\displaystyle S_n}$, since these two random variables have the same probability distribution.

Poissonized Plancherel measure

Plancherel measure is defined on ${\displaystyle S_n}$ for each integer ${\displaystyle n}$. In various studies of the asymptotic behavior of ${\displaystyle L(\sigma )}$ as ${\displaystyle n\to \mathrm{\infty }}$, it has proved useful to extend the measure to a measure, called the Poissonized Plancherel measure, on the set ${\displaystyle {\mathcal{P}}^{\ast }}$ of all integer partitions. For any ${\displaystyle \theta >0}$, the Poissonized Plancherel measure with parameter ${\displaystyle \theta }$ on the set ${\displaystyle {\mathcal{P}}^{\ast }}$ is defined by

${\displaystyle {\mu }_{\theta }(\lambda )=e^{-\theta }\frac{{\theta }^{|\lambda |}(f^{\lambda }{)}^2}{(|\lambda |!{)}^2},}$

for all ${\displaystyle \lambda \in {\mathcal{P}}^{\ast }}$.

Plancherel growth process

The Plancherel growth process is a random sequence of Young diagrams ${\displaystyle {\lambda }^{(1)}=(1),{\lambda }^{(2)},{\lambda }^{(3)},\dots ,}$ such that each ${\displaystyle {\lambda }^{(n)}}$ is a random Young diagram of order ${\displaystyle n}$ whose probability distribution is the nth Plancherel measure, and each successive ${\displaystyle {\lambda }^{(n)}}$ is obtained from its predecessor ${\displaystyle {\lambda }^{(n-1)}}$ by the addition of a single box, according to the transition probability

${\displaystyle p(\nu ,\lambda )=\mathbb{P}({\lambda }^{(n)}=\lambda |{\lambda }^{(n-1)}=\nu )=\frac{f^{\lambda }}{n{f}^{\nu }},}$

for any given Young diagrams ${\displaystyle \nu }$ and ${\displaystyle \lambda }$ of sizes n − 1 and n, respectively.

So, the Plancherel growth process can be viewed as a natural coupling of the different Plancherel measures of all the symmetric groups, or alternatively as a random walk on Young's lattice. It is not difficult to show that the probability distribution of ${\displaystyle {\lambda }^{(n)}}$ in this walk coincides with the Plancherel measure on ${\displaystyle S_n}$.

Compact groups

The Plancherel measure for compact groups is similar to that for finite groups, except that the measure need not be finite. The unitary dual is a discrete set of finite-dimensional representations, and the Plancherel measure of an irreducible finite-dimensional representation is proportional to its dimension.

Abelian groups

The unitary dual of a locally compact abelian group is another locally compact abelian group, and the Plancherel measure is proportional to the Haar measure of the dual group.

Semisimple Lie groups

The Plancherel measure for semisimple Lie groups was found by Harish-Chandra. The support is the set of tempered representations, and in particular not all unitary representations need occur in the support.