next up previous
Next: Bibliography Up: ON THE CONFLUENT HYPERGEOMETRIC Previous: ON THE CONFLUENT HYPERGEOMETRIC

Introduction

Let $\lambda=(\lambda _{0},\dots,\lambda _{l-1})$ be a partition of $n$, sometimes called a `Young diagram $\lambda'$ of weight $n$. Let $H\sb{\lambda}=J(\lambda\sb{0})\times\cdots\times J(\lambda\sb{l-1})\subset GL(n)$ be the associated maximal abelian subgroup with respect to $\lambda $, where $J(m)$ is the Jordan group of size $m$, i.e.,

\begin{displaymath}J(m)=\left\{\sum\sp{m-1}\sb{i=0} h\sb{i}\Lambda\sp{i}\,\vert\...
...n{\bf C}\sp{\times}, h\sb{1},\dots,h\sb{m-1}\in{\bf C}\right\},\end{displaymath}

where the $m \times m $ matrix $\Lambda$ is defined as

\begin{displaymath}\font\b=cmr10 scaled\magstep4\Lambda=\pmatrix{
0 & 1 & &\s...
...s & \cr
& {}&\ddots & 1\cr
\smash{\hbox{\b 0}}& & {} & 0 \cr
}\end{displaymath}

We define the biholomorphic map

\begin{eqnarray*}
\iota:H\sb{\lambda} & \longrightarrow& \prod_i ({\bf C}\sp{\ti...
...ots, h\sp{(l-1)}\sb{0},\dots, h\sp{(l-1)}\sb{\lambda\sb{l-1}-1})
\end{eqnarray*}



where $h=(h\sp{(0)},\dots, h\sp{(l-1)})$, $h\sp{(i)}=
\sum_{0\le k<\lambda\sb{i}} h\sp{(i)}\sb{k}\Lambda\sp{k}\in J(\lambda _i)$.

Let $
\alpha = (\alpha\sp{(0)} ,\dots , \alpha\sp{(l-1)} ),
\alpha^{(i)}:= (\alpha\sb{0}^{(i)} ,\dots , \alpha\sb{\lambda _i-1}^{(i)})\quad (0\le i\le l-1) $ be an $n$-tuple of complex numbers satisfying $\sum_{i=0}^{l-1}\alpha_0\sp{(i)}=-r.$ We define the character $\chi(\cdot\;;\;\alpha ) : \tilde{H}\sb{\lambda}\longrightarrow
{\bf C}\sp{\times}$ of the universal covering group $\tilde{H}_{\lambda }=\tilde{J}(\lambda _0)
\times\cdots\times\tilde{J}(\lambda _{l-1})$ of $H\sb{\lambda}$ by $\chi(h;\alpha )=\prod_{i=0}^{l-1}\chi(h^{(i)};\alpha ^{(i)})$, where

\begin{displaymath}\chi(h^{(i)};\alpha ^{(i)})=h_0^{\alpha _0^{(i)}}\exp\left[\s...
...ts, \frac {h^{(i)}_{\lambda _i-1}}{ h\sb{0}^{(i)}}\big)\right] \end{displaymath}

where $\theta\sb{j}$ are defined as the coefficients of the generating series

\begin{displaymath}\log(1+x\sb{1}T+x\sb{2}T\sp{2}+\cdots)=\sum\sp{\infty}\sb{j=0}\theta\sb{j}(x\sb{1},\dots,x\sb{j})T\sp{j}.\end{displaymath}

Recall that the hypergeometric function $\Phi(z;\alpha )$ of type $\lambda $ (see [7]) is a function defined by

\begin{displaymath}
\Phi(z;\alpha)=\int\sb{\Delta}\chi(\iota\sp{-1}(tz);\alpha)\cdot\omega\quad
\mbox{for} \quad z\in Z\sb{r,n}
\end{displaymath} (1)

where $ Z\sb{r,n} $ is the set of $r\times n$ complex matrices in general position (see [7]) with respect to $\lambda $, $\omega := \sum_{0\le i<r} (-1)\sp{i} t\sb{i} dt\sb{0}\wedge \cdots \wedge dt\sb{i-1}\wedge dt\sb
{i+1}\wedge\cdots\wedge dt\sb{r-1}$ and $\Delta$ is a twisted cycle in the $t$-space depending on $z$ and $\alpha $. Note that for $\lambda =(1,\dots,1)$, the hypergeometric functions of type $\lambda $ coincide with the general hypergeometric function defined in [5].

The set $Z_{r,n}$ admits an action of the group $GL(r)\times H_{\lambda }:$

\begin{eqnarray*}
GL(r)\times Z_{r,n}\times H_{\lambda }&\longrightarrow& Z_{r,n}\\
(g,z,h)\qquad &\longmapsto& gzh,
\end{eqnarray*}



under which $\Phi$ behaves as
$\displaystyle \Phi(gz;\alpha)$ $\textstyle =$ $\displaystyle ({\rm det}\, g)\sp{-1}\Phi(z;\alpha)\qquad g\in GL(r)$ (2)
$\displaystyle \Phi(zh\sb{\lambda};\alpha)$ $\textstyle =$ $\displaystyle \chi(h\sb{\lambda})\Phi(z;\alpha)\qquad h\sb{\lambda}\in H\sb{\lambda}.$ (3)

Furthermore, the function $\Phi$ admits another symmetry:
\begin{displaymath}
\Phi(zw\sb{\lambda};\alpha)=\Phi(z;\alpha {}\sp {t}w\sb{\lambda})\qquad w\sb{\lambda}\in W\sb{\lambda},
\end{displaymath} (4)

where $W\sb{\lambda}$ is an analogue of the Weyl group, see [8].

The hypergeometric functions $\Phi$ on $Z\sb{2,4}$ and $Z\sb{2,5}$ for various partitions $\lambda $ of 4 and 5 were investigated in the papers [7], [9] and [8]. It is known that the functions $\Phi$ are generalizations of Gauss', Kummer's, Bessel's, Hermite's, Airy's functions and the classical hypergeometric functions of two variables, i.e., $F\sb{1}, \Phi\sb{1}, \Phi\sb{2}, \Phi\sb{3}, G\sb{2}, \Gamma\sb{1}, \Gamma\sb{2}$ in Horn's list ([3]). In this paper, we study the hypergeometric functions of type $\lambda $ in two variables on the strata of the set $M(3,6)$ of $3\times 6$ complex matrices.We establish a classification of the functions in terms of the orbital decomposition of the set of strata and give some transformation formulae between some systems of differential equations satisfied by the functions $F\sb{2}, F\sb{3}, H\sb{2}, {\bf H}_3, {\bf H}_{11}, \Psi_1, \Xi_2.$

In Section 1 , we introduce a group $W\sb{\lambda^{(\nu)}}:=R_{\lambda^{(\nu)}} \times \kern -4pt \vert P\sb{\lambda^{(\nu)}}$ which is analogous to the (classical) Weyl group and discuss its properties in detail. In Section 2, we consider the action of $P\sb{\lambda^{(\nu)}}$ on the set $S_{\lambda ^{(\nu)}}$ of strata and obtain the orbital decomposition of $S_{\lambda ^{(\nu)}}$. In Section 3, we obtain suitable normal forms of the matrices in the strata relative to the action of $GL(3)\times H\sb{\lambda}$. Then we can reduce the hypergeometric function $\Phi$ into a function of two variables. In Section 4, we give relations between the classical special functions of hypergeometric type, i.e., Appell's $F\sb{2}$, $F\sb{3}$, $H\sb{2}$ and their confluence in Horn's list, and the hypergeometric functions of type $\lambda $ in two variables. In the last section, some transformation formulae for the systems of diffenentical equations are systematically deduced from the symmetries % latex2html id marker 380
$(\ref{equation:0.2})$- % latex2html id marker 382
$(\ref{equation:0.4})$ for the functions $\Phi$.

The author expresses his deep gratitude to Prof. Kazuo Okamoto, Prof. Hironobu Kimura and Prof. Katsunori Iwasaki for their valuable suggestions and kind help during the preparation of this paper.


next up previous
Next: Bibliography Up: ON THE CONFLUENT HYPERGEOMETRIC Previous: ON THE CONFLUENT HYPERGEOMETRIC
Nobuki Takayama 2002-09-18