Time Evolution of Averaged Limit Shapes of Random Multiple Young Diagrams
Akihito HORA · 2025
Read Time Evolution of Averaged Limit Shapes of Random Multiple Young Diagrams by Akihito HORA (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-07393.txt.
Introduction
Akihito Hora, writing alone in 2025, follows how random stacks of Young diagrams change over time. Young diagrams are staircase shapes made of boxes that encode symmetry data in algebra. Here several diagrams evolve together as boxes jump between them under a branching rule tied to wreath products, group constructions that mix a finite group with symmetric groups.
Under suitable scaling, averaged limit shapes emerge, smooth outlines that the random diagrams follow on average. Hora describes how those macroscopic shapes evolve in time using Voiculescu's R-transforms and free Lévy measures, tools from free probability theory. Both diffusive scaling, where motion spreads like heat, and a non-diffusive kind are covered. The microscopic dynamics even allow waiting times that are not exponential, a flexibility most such models lack. The work links random diagram growth to deep probability theory.
Catalog record
| CREATOR | Akihito HORA |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 52 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-07393
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Time Evolution of Averaged Limit Shapes of Random Multiple Young Diagrams Thanks: This work was supported by JSPS KAKENHI Grant Number JP22K03346 Thanks: Keywords: random multiple Young diagram, evolution of averaged limit shape, branching rule for wreath product groups, free cumulant of Kerov transition measure, infinite divisibility in the free sense. Akihito HORA Abstract The branching rule for the tower of wreath products of a finite group by the symmetric groups induces a stochastic process on the set of multiple Young diagrams through random transitions of boxes of the diagrams between one another. We observe dynamical multiple averaged limit shapes resulting from appropriate scaling limits, either diffusive or non-diffusive. We describe time evolution of the macroscopic multiple averaged limit shapes in terms of Voiculescu’s R R -transforms and free Lévy measures of corresponding Kerov transition measures. Our microscopic dynamics admits non-exponential pausing time distributions. 1 Introduction In the present note, we observe a certain time evolution of multi-interfaces as indicated in Figure 1 (right). Sufficiently zoomed in, each interface looks like a zigzag profile of a Young diagram. Moreover, the profiles fluctuate by exchanging boxes at random between one another. The random structure of our model comes from the branching rule for a tower of wreath product groups. Actually, a broad aim of our project is to study asymptotic behavior of the brancing rule through probability models. Let us settle the model precisely. Figure 1: Random multi-diagrams to multi-interfaces: the case of |\widehat{T}|=3 We consider the wreath product of a finite group T by the symmetric group \mathfrak{S}_{n} of degree n : \mathfrak{S}_{n}(T)=T\wr\mathfrak{S}_{n}=T^{n}\rtimes\mathfrak{S}_{n}. The set of conjugacy classes and the set of equivalence classes of the irreducible representations of T are denoted by [T] and \widehat{T} respectively. The symbol \mathbb{Y}_{n} denotes the set of Young diagrams with n boxes. Set \mathbb{Y}=\bigsqcup_{n=0}^{\infty}\mathbb{Y}_{n} where \mathbb{Y}_{0} consists of empty diagram \varnothing . The set of conjugacy classes [\mathfrak{S}_{n}(T)] and the set of equivalence classes of the irreducible representations (\mathfrak{S}_{n}(T))^{\wedge} of \mathfrak{S}_{n}(T) are parametrized by the |[T]|=|\widehat{T}| -tuples of Young diagrams with the numbers of boxes summing up to n , denoted by \mathbb{Y}_{n}([T]) and \mathbb{Y}_{n}(\widehat{T}) respectively: \displaystyle[\mathfrak{S}_{n}(T)]\cong\mathbb{Y}_{n}([T])=\bigl\{\rho=(\rho_{\theta})_{\theta\in[T]}\,\big|\,\rho_{\theta}\in\mathbb{Y},\;\sum_{\theta\in[T]}|\rho_{\theta}|=n\bigr\}, \displaystyle(\mathfrak{S}_{n}(T))^{\wedge}\cong\mathbb{Y}_{n}(\widehat{T})=\bigl\{\lambda=(\lambda^{\zeta})_{\zeta\in\widehat{T}}\,\big|\,\lambda^{\zeta}\in\mathbb{Y},\;\sum_{\zeta\in\widehat{T}}|\lambda^{\zeta}|=n\bigr\} where |\cdot| denotes the number of boxes of a Young diagram. For k,n\in\mathbb{N} such that k<n , the canonical inclusion \iota_{n,k}:\mathfrak{S}_{k}(T)\longrightarrow\mathfrak{S}_{n}(T) \displaystyle x\in T^{k}\longmapsto(x,e_{T},\cdots,e_{T})\in T^{k}\times T^{n-k}=T^{n}, \displaystyle\sigma\in\mathfrak{S}_{k}\longmapsto\sigma(k+1)\cdots(n)\in\mathfrak{S}_{n} induces the following inclusion, which is denoted again by \iota_{n,k}:\mathbb{Y}_{k}([T])\longrightarrow\mathbb{Y}_{n}([T]) . For \rho=(\rho_{\theta})_{\theta\in[T]}\in\mathbb{Y}_{k}([T]) , \iota_{n,k}\rho=(\tilde{\rho}_{\theta})_{\theta\in[T]}\in\mathbb{Y}_{n}([T])\qquad\text{where}\quad\tilde{\rho}_{\theta}=\begin{cases}\rho_{\{e_{T}\}\sqcup(1^{n-k})},&\theta=\{e_{T}\}\\ \rho_{\theta},&\text{otherwise}.\end{cases} For simplicity we often write as \mathfrak{S}_{k}(T) instead of \iota_{n,k}\mathfrak{S}_{k}(T) to have \mathfrak{S}_{k}(T)\subset\mathfrak{S}_{n}(T) . For \lambda=(\lambda^{\zeta})_{\zeta\in\widehat{T}}\in\mathbb{Y}_{n}(\widehat{T}) and \nu=(\nu^{\zeta})_{\zeta\in\widehat{T}}\in\mathbb{Y}_{n-1}(\widehat{T}) , we write as \nu\nearrow\lambda or \lambda\searrow\nu when \lambda is obtained by adding a box at one entry of \nu . The uniquely determined entry label is then denoted by \zeta_{\nu\lambda}\in\widehat{T} . Let \pi^{\lambda} be an irreducible unitary representation of \mathfrak{S}_{n}(T) corresponding to \lambda\in\mathbb{Y}_{n}(\widehat{T}) . The dimension of \pi^{\lambda} is given by \dim\lambda=\frac{n!}{\prod_{\zeta\in\widehat{T}}|\lambda^{\zeta}|!}\prod_{\zeta\in\widehat{T}}(\dim\zeta)^{|\lambda^{\zeta}|}\dim\lambda^{\zeta}. (1.1) We have irreducible decomposition with dimension counting: \mathrm{Res}^{\mathfrak{S}_{n}(T)}_{\mathfrak{S}_{n-1}(T)}\pi^{\lambda}\cong\!\!\bigoplus_{\nu\in\mathbb{Y}_{n-1}(\widehat{T}):\,\nu\nearrow\lambda}[\dim\zeta_{\nu\lambda}]\pi^{\nu},\quad\dim\lambda=\!\!\sum_{\nu\in\mathbb{Y}_{n-1}(\widehat{T}):\,\nu\nearrow\lambda}\dim\zeta_{\nu\lambda}\dim\nu (1.2) and reciprocally, \mathrm{Ind}^{\mathfrak{S}_{n}(T)}_{\mathfrak{S}_{n-1}(T)}\pi^{\nu}\cong\!\!\bigoplus_{\mu\in\mathbb{Y}_{n}(\widehat{T}):\,\nu\nearrow\mu}[\dim\zeta_{\nu\mu}]\pi^{\mu},\quad n|T|\dim\nu=\!\!\sum_{\mu\in\mathbb{Y}_{n}(\widehat{T}):\,\nu\nearrow\mu}\dim\zeta_{\nu\mu}\dim\mu. (1.3) ( 1.2 ) and ( 1.3 ) produce two stochastic matrices by P^{\downarrow}_{\lambda\nu}=\frac{\dim\zeta_{\nu\lambda}\dim\nu}{\dim\lambda},\qquad P^{\uparrow}_{\nu\mu}=\frac{\dim\zeta_{\nu\mu}\dim\mu}{n|T|\dim\nu} and thereby a stochastic matrix of degree |\mathbb{Y}_{n}(\widehat{T})| P=P^{(n)}=P^{\downarrow}P^{\uparrow}. (1.4) More explicitly, the entries of P of ( 1.4 ) are given by P_{\lambda\mu}=\begin{cases}\dfrac{1}{n}\,\dfrac{|\lambda^{\zeta}|}{|\mu^{\eta}|}\,\dfrac{\dim\mu^{\zeta}\dim\mu^{\eta}}{\dim\lambda^{\zeta}\dim\lambda^{\eta}}\,\dfrac{(\dim\eta)^{2}}{|T|},&\exists\zeta,\eta\text{ s.t. }\lambda^{\zeta}\searrow\mu^{\zeta},\ \lambda^{\eta}\nearrow\mu^{\eta}\\ \dfrac{1}{n}\,\dfrac{\dim\mu^{\zeta}}{\dim\lambda^{\zeta}}\,\dfrac{(\dim\zeta)^{2}}{|T|},&\lambda\neq\mu;\ \forall\zeta,\;|\lambda^{\zeta}|=|\mu^{\zeta}|\\ \dfrac{1}{n|T|}\sum_{\zeta}(\dim\zeta)^{2}\cdot\#\{\text{peaks of }\lambda^{\zeta}\},&\lambda=\mu\\ \ 0,&\text{otherwise}.\end{cases} The Markov chain induced by ( 1.4 ), called Res-Ind chain on \mathbb{Y}_{n}(\widehat{T}) , is symmetric (i.e. satisfies detailed balance) with respect to the Plancherel measure M_{\mathrm{Pl}}(\lambda)=\frac{(\dim\lambda)^{2}}{n!|T|^{n}},\qquad M_{\mathrm{Pl}}(\lambda)P_{\lambda\mu}=M_{\mathrm{Pl}}(\mu)P_{\mu\lambda},\qquad\lambda,\mu\in\mathbb{Y}_{n}(\widehat{T}). The procedure of defining a Res-Ind chain works well for a general pair (G,H) of a finite group and its subgroup. Decomposition of \mathrm{Ind}\circ\mathrm{Res}\,\pi^{\lambda} for wreath products was considered in [ 17 ] . Res-Ind chains were introduced by Fulman to apply to the Stein method for algebraic central limit theorems in [ 5 ] , [ 6 ] . See [ 3 ] for an application to infinite-dimensional diffusions. We have treated Res-Ind chains in a series of works [ 10 ] , [ 11 ] , [ 12 ] and [ 13 ] preceding the present note. Iteration of ( 1.2 ) and ( 1.3 ) along n produces the branching graph for the towers of wreath product groups. Our series of works [ 9 ] , [ 15 ] and [ 14 ] treat related harmonic analysis where asymptotic analysis is developed in a different scaling regime from this note. Interesting probability models on such a branching graph are studied in [ 19 ] . A static model on fluctuations for growing multiple Young diagrams is studied in [ 18 ] . Substituting a counting process for discrete time k of Res-Ind chain (Z_{k}^{(n)})_{k\in\{0,1,2,\dots\}} , we obtain a continuous time process on \mathbb{Y}_{n}(\widehat{T}) . For that purpose, let \psi be a pausing (or holding) time distribution, which is a probability on (0,\infty) . Take an IID sequence (\epsilon_{j})_{j\in\mathbb{N}} , each obeying \psi , and set N_{s}=\begin{cases}0,&s<\epsilon_{1}\\ j,&\epsilon_{1}+\cdots+\epsilon_{j}\leqq s<\epsilon_{1}+\cdots+\epsilon_{j+1},\end{cases}\qquad N_{0}=0\ \text{a.s.} The continuous time process (X_{s}^{(n)})_{s\geqq 0} obtained by X_{s}^{(n)}=Z_{N_{s}}^{(n)},\qquad s\geqq 0 (1.5) describes our microscopic random dynamics. The distribution at time s is expressed as \mathbb{P}(X_{s}^{(n)}=\mu\,|\,X_{0}^{(n)}=\lambda)=\sum_{j=0}^{\infty}(P^{(n)j})_{\lambda\mu}\int_{[0,s]}\psi\bigl((s-u,\infty)\bigr)\psi^{\ast j}(du),\qquad\lambda,\mu\in\mathbb{Y}_{n}(\widehat{T}) (1.6) where \psi^{\ast j} is the j th convolution of \psi . See [ 12 ] , [ 13 ] for ( 1.6 ) 1 1 1 The assumption of \psi((0,\infty))>0 in [ 12 ] and [ 13 ] is insufficient and replaced by \psi(\{0\})=0 for N_{0}=0 a.s. . Transfer from microscopic random multi-diagrams to macroscopic multi-interfaces as Figure 1 is formulated through appropriate space–time scaling limits. In the spatial direction, 1/\sqrt{n} rescale is nice for the limit shape of Young diagrams. We set (\lambda^{\zeta})^{\sqrt{n}}(x)=\frac{1}{\sqrt{n}}\lambda^{\zeta}(\sqrt{n}x),\qquad\lambda=(\lambda^{\zeta})_{\zeta\in\widehat{T}}\in\mathbb{Y}_{n}(\widehat{T}) (1.7) where y=\lambda^{\zeta}(x) is the profile of \lambda^{\zeta} displayed in the xy -plane. In the temporal direction, letting t be a macroscopic time, we set s=t\tau_{n} where (\tau_{n}) is a positive sequence such that \lim_{n\to\infty}\tau_{n}=\infty . Later the diverging order \tau_{n} will be taken in several ways. Compared with spatial 1/\sqrt{n} rescale, the case of \tau_{n}=n gives a diffusive limit. In our project, we investigate the behavior of random multi-diagrams \bigl((X_{t\tau_{n}}^{(n)\,\zeta})^{\sqrt{n}}\bigr)_{\zeta\in\widehat{T}} (1.8) as n\to\infty and time evotution (i.e. concrete t -dependence) of limiting objects by setting appropriate assumptions on initial ensembles (i.e. distributions of X_{0}^{(n)} ) and pausing time distributions. In this note, we derive the time evolution of some averaged quantities for ( 1.8 ) by applying harmonic analysis of wreath product groups. The results for averaged limit shapes in the case of stable pausing time distributions are new, as continuation from [ 12 ] , even if they are restricted to trivial T . A more detailed aspect of dymanical multi-interfaces as a concentration phenomenon (i.e. law of large numbers) will be discussed in a separate note. The subsequent sections are organized as follows. The next §2 is devoted to reviewing some necessary facts including the character formula for wreath product groups. In §3, we state evolution of the averaged areas (i.e. numbers of boxes) of rescaled multi-diagrams, which are the simplest macroscopic quantities resulting from interaction between multi-diagrams. In §4, we investigate t -dependence of the transition measures to fully describe the averaged dynamical limit shapes. Computation of their free cumulants is essential. Examples of typical ensembles are constructed from characters of the infinite wreath product \mathfrak{S}_{\infty}(T) in §5. 2 Preliminaries 2.1 Character formula for wreath product groups Character formula of a