On some quadratic algebras
Extracted source text ยท arXiv q-alg/9705003
%====================================================================
%
% On some quadratic algebras
%
% Anatol N.Kirillov, April, 1997, last May 12
%
% LATEX
%=====================================================================
\documentclass[12pt]{article}
\usepackage{amssymb,latexsym}
\setlength{\topmargin}{0.0in}
\setlength{\textheight}{21.5cm}
\setlength{\evensidemargin}{0.35in}
\setlength{\oddsidemargin}{0.35in}
\setlength{\headsep}{0.1cm}
\setlength{\textwidth}{15.0cm}
\setlength{\parindent}{0.6cm}
\setlength{\unitlength}{1mm}
\renewcommand{\theequation}{\arabic{section}.\arabic{equation}}
\def\neweq{\setcounter{equation}{0}}
\newtheorem{theorem}{Theorem}[section]
\newtheorem{pr}[theorem]{Proposition}
\newtheorem{cor}[theorem]{Corollary}
\newtheorem{de}[theorem]{Definition}
\newtheorem{con}[theorem]{Conjecture}
\newtheorem{rem}[theorem]{Remark}
\newtheorem{lem}[theorem]{Lemma}
\newtheorem{fact}[theorem]{Fact}
\newtheorem{example}[theorem]{Example}
\newtheorem{axiom}[theorem]{Axiom}
\newtheorem{property}[theorem]{Property}
\newtheorem{problem}[theorem]{Problem}
\newtheorem{ex}[theorem]{Example}
\def\proof{\smallskip\noindent {\it Proof --- \ }}
\def\endproof{\hfill$\square$\medskip}
\def\endproofmath{\quad\square}
\font\germ=eufm10
\def\Eg{\hbox{\germ E}}
\def\pa{\partial}
\def\S{\mathfrak{S}}
\def\Q{\mathbb{Q}}
\def\HQ{{\cal HQ}}
\def\O{\Omega}
\def\bO{\overline{\Omega}}
\def\bs{\overline{\sigma}}
\def\CH{{\rm CH}}
\def\QH{{\rm QH}}
\def\l{\ell}
\def\A{{\cal A}}
\def\B{{\cal B}}
\def\D{{\cal D}}
\def\E{{\cal E}}
\def\G{{\cal G}}
\def\H{{\cal H}}
\def\I{{\cal I}}
\def\K{{\cal K}}
\def\L{{\cal L}}
\def\M{{\cal M}}
\def\P{{\cal P}}
\def\R{{\cal R}}
\def\T{{\cal T}}
\def\Eqn{{{\cal E}^q_n}}
\def\point{\textbf{.}}
\def\s{\hbox{\germ S}}
\def\<{\langle}
\def\>{\rangle}
\def\tE{\widetilde E}
\def\X{{\mathcal X}}
\def\tX{\widetilde \X}
\def\ttX{{\buildrel\approx\over \X}}
\def\zplusq{{\mathbb{Z}_+[q]}}
\def\ibid{{---------}}
\def\wnot{w_\mathrm{o}}
\def\Z{\mathbb{Z}}
\def\C{\mathbb{C}}
\def\wt{\widetilde}
\def\wh{\widehat}
\def\Enp{\E_n^+}
\def\dunkl{\theta}
\def\Dunkl{\Theta}
\def\wsum{\sum^{\widetilde{\ \quad\ }}}
\def\smallwsum{\displaystyle\sum^{\widetilde{\ \ \ }}}
\def\ds{\displaystyle}
\def\qed{\hfill$\vrule height 2.5mm width 2.5mm depth 0mm$}
\title{On some quadratic algebras
\date{ }
\author{Anatol N. Kirillov \\~ \\
{\small {\it CRM, University of Montreal}} \\
{\small {\it C.P. 6128, Succursale A, Montreal (Quebec), H3C 3J7,
Canada}}\\
{\small {\it and}}
\\
{\small {\it Steklov Mathematical Institute,}} \\
{\small {\it Fontanka 27, St.Petersburg, 191011, Russia}}}
}
\begin{document}
\maketitle
\begin{abstract}
We study some quadratic
algebras which are appeared in the low--dimen\-sio\-nal topology and Schubert
calculus.
We introduce the Jucys--Murphy elements in the braid algebra and in the pure
braid group, as well as the Dunkl elements in the extended affine braid group.
Relationships between the Dunkl elements, Dunkl operators and
Jucys--Murphy elements are described.
\end{abstract}
\section{Introduction}
\label{sec:intro}
\neweq
In this paper we study some quadratic algebras which are naturally
appeared in the low--dimensional topology [B-N], [Dr2], [Ko2] and Schubert
calculus [FK], and investigate some of their properties. We define a quadratic
algebra $\G_n$, which is a further generalization of the quadratic algebras
$\E_n$ and $\E_n^t$ introduced in [FK],
Sections~2 and 15. Our main idea is to apply the results obtained for the
braid algebra $\B_n$, [Dr2], [Ko2], to the algebra $\G_n$. The main
observation is that some results which are well--known for the braid
algebra $\B_n$, can be
(re)proven for the quadratic algebra $\G_n$. For example,
it happens that the quadratic algebra $\G_n$ and the braid algebra $\B_n$
have the same Hilbert series
$$H(\B_n;t)=H(\G_n;t)=\frac{1}{(1-t)(1-2t)\cdots (1-(n-1)t)},
$$
%\vbox{\hskip -0.6cm
and have the commutative subalgebras $\K_n\subset\B_n$ and
$\H_n\subset\G_n$ which are both isomorphic to the ring of polynomials.
The algebra $\K_n$ is generated by the
Jucys--Murphy elements, whereas the algebra $\H_n$ is generated by the Dunkl
ones. Both algebras $\G_n$ and $\B_n$ have the additive bases consisting
of all monomials in normal form (Theorem~2.4 and Corollary~4.3; cf. [L]). %}
We expect also that the certain quotients $\E_n^0$ and $\B_n^0$ of the
quadratic algebras $\G_n$ and $\B_n$ respectively, have the same Hilbert
polynomials as well (see Section~9 for details).
One of the main goals of this paper is to describe the
relationships between the Dunkl and Jucys--Murphy elements, and to compute
the dimensions of
homogeneous component of degree $\le 6$ of the quadratic algebra $\E_n^0$.
The structure of the paper is following:
In Section~2 we define the braid algebra $\B_n$ as the infinitesimal
deformation of the pure braid group $P_n$ (cf. T.~Kohno [Ko2]).
%It is interesting to note that
The algebra $\B_n$ and its completion $\wh\B_n$ have many important
applications to the low--dimensional topology, see e.g. [B-N], [Dr2], [F],
[Ko3], [L].
In Section~3 we define the Jucys--Murphy elements $d_k$ in the braid
algebra $\B_n$, and the multiplicative Jucys--Murphy elements $D_k$ in
the pure braid group $P_n$. Follow to A.~Ram, [Ra], we prove that the
Jucys--Murphy element $d_k\in\B_n$ is the quasi--classical limit of the
element $D_k\in P_n$. We prove also, that the infinitesimal deformation
of the multiplicative Jucys--Murphy element $D_k$ coincides with the
element $d_k$.
In Section~4 we define the
quadratic algebra $\G_n$ and compute the Hilbert series for this algebra.
In Section~5 we define the Dunkl elements $\theta_j$ in the quadratic algebra
$\G_n$ and describe a commutative subalgebra generated by $\theta_j$,
$1\le j\le n$.
In Section~6, which contains one of the main results of this paper,
we study a relationship between the Jucys--Murphy elements
and the Dunkl elements. We define the dual Dunkl elements $Y_1^*,\ldots
Y_n^*$ as certain elements in the extended affine braid group $\wt B_n$, and
prove that under a natural homomorphism $\pi :\wt B_n\to B_n$, the dual
Dunkl element $Y_k^*$ maps to the multiplicative Jucys--Murphy element
$D_k$. We explain also a connection between Dunkl element $Y_k\in\wt B_n$
and the Dunkl operator ${\bf Y}_k\in\D_{q,x}[W]$, where $\D_{q,x}[W]$ is
the algebra of $q$--difference operators with permutations.
In Section~7 we study the relations in the algebra $\E_n^t$. In
Section~8 we use the properties of "normal basis" introduced in Section~4,
to study the
quotient algebra $\E_n^0$, and compute the dimensions of homogeneous
components $\E_{n,k}^0$,
$1\le k\le 6$, and also the Hilbert polynomial $H(\E_4^0;t)$.
The last polynomial was also computed, using computer, by J.-E.~Ross [Ro],
who computed the Hilbert polynomial $H(\E_5^0;t)$ as well.
In Section~9 we consider a certain quotient $\B_n^0$ of the braid algebra
$\B_n$, and make a conjecture that the quadratic algebra $\E_n^0$ and
$\B_n^0$ have the same Hilbert polynomials. In particular, we expect that
the algebra $\B_n^0$ (as well as $\E_n^0$) has a finite dimension.
In Section~10 we study the commutative quadratic algebra $\A_n^t$ (denote by
$\E C_n$ in [FK], Section~4.3), which is a commutative quotient of $\E_n^t$
(see Definition~\ref{d7.1}), and explain some details which were omitted in
[FK].
\medskip
\textsc{Acknowledgments.}
This work was done during my stay at the CRM, University of Montreal. I would
like to thank all my colleagues for discussions and support. I would like
also to thank M.~Noumi and A.N.~Varchenko for fruitful and stimulating
discussions. My
special thanks to Arun Ram, who explained me during my stay at the
University of Wisconsin (1995) the importance of the Jucys--Murphy
elements in the representation theory, and N.A.~Liskova for an
inestimable help on different stages of this paper and constructive
criticism.
\section{Quadratic algebra $\B_n$}
\label{sec:quadB}
\neweq
We start with consideration of the quadratic algebra $\B_n$ which is the
infinitesimal deformation of the pure braid group $P_n$, [Ko2]. This
quadratic algebra is well--known as the algebra of (finite order)
Vassiliev invariants of braids, [Ko3], [F], [B-N]. The completion
$\wh\B_n$ of $\B_n$ with respect to the grading was considered in the
papers of V.~Drinfeld, [Dr1], [Dr2], in his study of quasi--Hopf algebras,
and in the papers of T.~Kohno, [Ko1]--[Ko3], in his study of monodromy
representations of braid groups.
\begin{de}\label{d3.1}{\rm([Ko2], [Dr2])}
Define the braid algebra $\B_n$ as the quadratic algebra
(say over $\Z$) with generators $X_{ij}$, $1\le i<j\le n$, which satisfy
the following relations
\begin{equation}\label{3.1}
1)~~X_{ij}\cdot X_{jk}-X_{jk}\cdot X_{ij}=X_{ik}\cdot X_{ij}-
X_{ij}\cdot X_{ik}=X_{jk}\cdot X_{ik}-X_{ik}\cdot X_{jk};
\end{equation}
\begin{equation}\label{3.2}
2)~~X_{ij}\cdot X_{kl}=X_{kl}\cdot X_{ij},~~{\rm if~~all}~~i,j,k,l
~~{\rm are~~distinct}.~~~~~~~~~~~~~~~~~~~
\end{equation}
\end{de}
\begin{rem} {\rm The algebra $\B_n$ is denoted by $\P_n$ in [L], and as
$A^n$ in [Ko1]--[Ko3].}
\end{rem}
The algebra $\B_n$ has a natural structure of a cocommutative Hopf
algebra. Namely, a comultiplication $\Delta
:~\B_n\to\B_n\otimes\B_n$, antipode $S:~\B_n\to\B_n$, and counit
$\epsilon :~A\to\Z$, can be define as follows
\begin{eqnarray*}
\Delta (X_{ij})&=&1\otimes X_{ij}+X_{ij}\otimes 1,\\
S(X_{ij})&=&-X_{ij},\\
\epsilon (X_{ij})&=&0,\ \ \epsilon (1)=1.
\end{eqnarray*}
Let us explain briefly an origin of the relations
(\ref{3.1})--(\ref{3.2}). For more details and proofs see [Dr1], [Dr2],
[F], [Ko1]--[Ko3], [L]. Let us denote by
$$\M=\C^n\setminus\bigcup_{1\le i<j\le n}\{ z_i=z_j\} ,
$$
the configuration space of $n$ distinct ordered points in $\C$. It is
well--known that the fundamental group $\pi_1(\M )$ of the space $\M$
coincides with the pure braid group $P_n$:
$$\pi_1(\M )=P_n.
$$
Let
$$w_{ij}=w_{ji}=\frac{1}{2\pi\sqrt -1}d\log (z_i-z_j)
$$
be closed 1--form on $\M$.
Then, $\{ w_{ij}~|~1\le i<j\le n\}$ represents a basis for $H^1(\M )$,
and the relations among $w_{ij}$, $1\le i<j\le n$, are generated by the
Arnold relations
$$w_{ij}\wedge w_{jk}+w_{jk}\wedge w_{ik}+w_{ik}\wedge w_{ij}=0
$$
for $i<j<k$ (see, e.g. [A]).
Let $\{ X_{ij}=X_{ji}~|~1\le i<j\le n\}$ be a set of non--commutative
variables, and consider a formal connection
\begin{equation}
\Omega =\sum_{1\le i<j\le n}w_{ij}X_{ij}. \label{4.4}
\end{equation}
\begin{lem}\label{l3.1} {\rm (T.~Kohno)} The connection $\Omega$ is
integrable if and only if the following conditions are satisfied
\begin{eqnarray*}
&& [X_{ij},X_{ik}+X_{jk}]=[X_{ij}+X_{ik}, X_{jk}]=0, \ {\rm for} \
i<j<k,\\
&&[X_{ij},X_{kl}]=0, \ {\rm for~ distinct} \ i,j,k,l.
\end{eqnarray*}
\end{lem}
Let us remind that $[a,b]=ab-ba$ is the usual commutator.
It is clear that these conditions are equivalent to the relations
(\ref{3.1})--(\ref{3.2}). A proof of Lemma~\ref{l3.1} follows from the
observations that $d\Omega =0$, and the integrability
condition for the connection $\Omega$ is equivalent to the following one:
$\Omega\wedge\Omega =0$. Notice that the connection (\ref{4.4}) is the
formal version of the so--called Knizhnik--Zamolodchikov connection in
conformal field theory.
The relations (\ref{3.1}) and (\ref{3.2}) can be also interpreted [Ch2],
[Dr2], [Ko2], as the self--consistency conditions
$$\frac{\partial A_j}{\partial z_i}-\frac{\partial A_i}{\partial z_j}
=[A_i,A_j], \ \ 1\le i,j\le n,
$$
of the Knizhnik--Zamolodchikov (system of) equation(s)
$$\frac{\partial\Phi}{\partial z_i}=kA_i\Phi, \ \ 1\le i\le n,
$$
where $A_i=\displaystyle\sum_{j\ne i}\frac{X_{ij}}{z_i-z_j}$, and a
function $\Phi :=\Phi (z)$, $z=(z_1,\ldots ,z_n)$, takes the values in a
certain algebra (say over $\C((z))$) generated by the elements $X_{ij}$.
Below we formulate the basic properties of the quadratic algebra $\B_n$.
\begin{theorem}\label{t3.1} {\rm ([Ko1], [Ko2])}
The Hilbert series of the algebra $\B_n$ is given by
\begin{equation}\label{3.3}
H(\B_n;t)=\frac{1}{(1-t)(1-2t)\cdots (1-(n-1)t)}.
\end{equation}
\end{theorem}
\begin{cor}\label{c3.2} {\rm ([Dr1], [L])} A linear basis in the algebra $\B_n$
is given
by the following set of noncommutative monomials
\begin{equation}
Z_2\cdot\ldots\cdot Z_n, \label{4.5}
\end{equation}
where
\begin{eqnarray*}
Z_2 & = & \{ 1,X_{12}\},\\
& \cdots & \\
Z_k & = & \{ 1\}\cup\{ X_{i_1k}\ldots X_{i_lk}~|~1\le i_j<k,~1\le j\le
l\}, \ \ 2\le k\le n.
\end{eqnarray*}
\end{cor}
A monomial in $X_{ij}$'s is called to be in {\it normal form} if it
belongs to the set (\ref{4.5}).
\begin{rem} {\rm Corollary \ref{c3.2} was stated without a proof in [Dr1].
The formula (\ref{3.3}) was obtained in [Ko1] by constructing a free
resolution of $\C$ as a trivial $\B_n$--module. Also, the formula
(\ref{3.3}) is a consequence of the fact that $\B_n$ is a semi-direct
product of free associative algebras as shown in [Dr2]; see also [L].
Note, that the relations (\ref{3.1}) can be interpreted as the horizontal
version of the 4--term relations in the theory of Vassiliev link
invariants.}
\end{rem}
Let us consider a completion $\widehat\B_n$ of the algebra
$\B_n\otimes_{\Z}\C$ with respect to the powers of the ideal $\I=(X_{ij})$
generated by $X_{ij}$, $1\le i<j\le n$. More precisely, let us denote by
$\C\langle X_{ij}\rangle$ the ring of non--commutative formal power
series with indeterminates $X_{ij}$, $1\le i<j\le n$, and let $J$ be an
ideal generated by the relations (\ref{3.1}) and (\ref{3.2}). Then $\widehat
\B_n=\C\langle X_{ij}\rangle /J$. It is well--known [Ko1], that
there exists an isomorphism of complete Hopf algebras
$$\widehat{\C [P_n]}\cong \widehat\B_n,
$$
where $\widehat{\C [P_n]}$ stands for the completion of the group ring
of the pure braid group $P_n$ with respect to the powers of the
augmentation ideal. %For further details see [Ko1], [Ko2], [Dr2], [B-N].
\section{Jucys--Murphy elements}
\label{sec:Juc}
\neweq
\begin{de}\label{d4.1} The Jucys--Murphy elements $d_j$, for $j=2,\ldots
,n$, in the quadratic algebra $\B_n$ are defined by
\begin{equation}\label{4.1}
d_j=\sum_{1\le i<j}X_{ij}.
\end{equation}
\end{de}
It is clear that $d_i$ is a primitive element, i.e.
$\Delta (d_i)=1\otimes d_i+d_i\otimes 1$.
\begin{lem}\label{l4.2} Relations $[X_{ik}+X_{jk},X_{ij}]=0$ for $i<j<k$,
together with commutativity relations (\ref{3.2}), imply that the
Jucys--Murphy elements $d_j$ commute pairwise.
\end{lem}
{\it Proof.} For $j<l$,
$$[d_j,d_l]=\sum_{1\le i<j, \ 1\le k<l}[X_{ij},X_{kl}]=
\sum_{1\le i<j<l}\left\{ [X_{ij},X_{il}]+[X_{ij},X_{jl}]\right\} =0.
$$
\qed
Let us assume that additionally to the relations (\ref{3.1}) and
(\ref{3.2}) the following relations are satisfied
\begin{equation}
X_{ij}^2=1,~~~1\le i<j\le n. \label{4.2}
\end{equation}
Then a map $(i<j)$
$$X_{ij}\to (i,j),
$$
where $(i,j)\in S_n$ is the transposition that interchanges $i$ and $j$
and fixes each $k\ne i,j$, defines a representation of the relations
(\ref{3.1}), (\ref{3.2}) and (\ref{4.2}). In other words, there exists a
homomorphism $p$ from the braid algebra $\B_n$ to the group ring of the
symmetric group, $p:~\B_n\to\Z [S_n]$, such that $p(X_{ij})=(i,j)$.
Under this homomorphism $p$ the
element $d_j$ maps to the Jucys--Murphy element in the group ring
of the symmetric group $S_n$ (see, e.g. [J],
[Mu], [Ra]). A proof follows from the following
relations between the transpositions in the symmetric group $S_n$: if $1\le
i<j<k\le n$, then
$$(ij)(ik)=(jk)(ij)=(ik)(jk),
$$
$$(ij)(jk)=(ik)(ij)=(jk)(ik).
$$
\begin{theorem}\label{t4.1} Let $\K_n$ be a commutative subalgebra of
$\B_n$ generated by the Jucys--Murphy elements $d_j$, $j=2,\ldots ,n$.
Then a map $d_j\mapsto x_{j-1}$ defines an isomorphism
$$\K_n\cong\Z [x_1,\ldots ,x_{n-1}].
$$
\end{theorem}
There exists a multiplicative analog, denoted by $D_k$, of the
Jucys--Murphy element $d_k$.
Our construction of the elements $D_k$ follows to [Ra].
Let $g_i$, $1\le i\le n-1$, be the standard generators of the braid group
$B_n$. Thus, the generators $g_i$ satisfy the following relations
\begin{eqnarray}
&& g_ig_j=g_jg_i, \ \ {\rm if} \ \ |i-j|\ge 2, \label{4.4'}\\
&&g_ig_{i+1}g_i=g_{i+1}g_ig_{i+1}, \ \ 1\le i\le n-2. \label{4.5'}
\end{eqnarray}
Now let us define the multiplicative Jucys--Murphy elements $D_k$:
\begin{de} {\rm (cf. [Ra], (3.16))} The multiplicative Jucys--Murphy
elements $D_k$ are defined by the following formulae
\begin{equation}
D_2=g_1^2; \
D_k=g_{k-1}g_{k-2}\cdots g_2g_1^2g_2\cdots g_{k-2}g_{k-1}, \ 3\le k\le n.
\label{4.6}
\end{equation}
\end{de}
\begin{lem}\label{l4.3} The elements $D_k$ commute pairwise.
\end{lem}
{\it Proof.} First of all,
$$D_2D_3=g_1^2g_2g_1^2g_2=g_1g_2g_1g_2g_1g_2=g_2g_1^2g_2g_1^2=D_3D_2.
$$
Now, using induction, we have ($3\le k\le n-1$)
\begin{eqnarray*}
D_kD_{k+1}&=& g_{k-1}D_{k-1}g_{k-1}g_kg_{k-1}D_{k-1}g_{k-1}g_k=
g_{k-1}g_kD_{k-1}g_{k-1}D_{k-1}g_kg_{k-1}g_k\\
&=& g_{k-1}g_kD_{k-1}g_{k-1}D_{k-1}g_{k-1}g_kg_{k-1}=
g_{k-1}g_kD_{k-1}D_kg_kg_{k-1}\\
&=& g_{k-1}g_kD_kD_{k-1}g_kg_{k-1}=D_{k+1}D_k.
\end{eqnarray*}
Finally, if $l-k\ge 2$, then
\begin{eqnarray*}
D_kD_l&=& D_kg_{l-1}\cdots g_{k+1}D_{k+1}g_{k+1}\cdots g_{l-1}=
g_{l-1}\cdots g_{k+1}D_kD_{k+1}g_{k+1}\cdots g_{l-1}\\
&=& g_{l-1}\cdots g_{k+1}D_{k+1}D_kg_{k+1}\cdots g_{l-1}=D_lD_k.
\end{eqnarray*}
\qed
The elements $D_k$, $2\le k\le n$,
generate a commutative subgroup $K_n$ in the braid group $B_n$. Now we
are going to consider the following two problems: what is the
infinitesimal deformation of the commutative subgroup $K_n$, and what is
the quasi--classical limit
\begin{equation}
\Z [B_n]\to H_n(q)\to\Z [S_n] \label{4.3}
\end{equation}
of the multiplicative Jucys--Murphy elements $D_k$? In the
quasi--classical limit (\ref{4.3}) the
group ring $\Z [B_n]$ of the braid group $B_n$ is degenerated at first to
the Iwahori--Hecke algebra $H_n(q)$, and then to the group ring $\Z
[S_n]$ of the symmetric group $S_n$.
\begin{pr}\label{p4.1} The quasi--classical limit (\ref{4.3}) of the
multiplicative Jucys--Murphy element $D_k\in B_n$ is equal to the
Jucys--Murphy element $d_k\in\Z [S_n]$.
\end{pr}
{\it Proof.}
Let us compute the quasi--classical limit (\ref{4.3}) of the element
$D_k$. The first step is to consider $D_k$ as an element of the
Iwahori--Hecke algebra $H_n(q)$. In other words, we have to add to
(\ref{4.4'}) and (\ref{4.5'}) the new relations
\begin{equation}
g_i^2=(q^{\frac{1}{2}}-q^{-\frac{1}{2}})g_i+1, \ \ 1\le i\le n-1.
\label{4.7}
\end{equation}
The next step is to consider the limit
$$\lim_{q\to 1}~\frac{D_k-1}{q^{\frac{1}{2}}-q^{-\frac{1}{2}}}.
$$
For this goal it is enough to compute $\ds\frac{dD_k}{dq}\vert_{q=1}$.
This can be done using the formula
$$\frac{dg_i}{dq}=\frac{(q^{\frac{1}{2}}-q^{-\frac{1}{2}})g_i+
2}{2q(q^{\frac{1}{2}}+q^{-\frac{1}{2}})}.
$$
Thus,
\begin{eqnarray*}
\frac{dD_k}{dq}&=& \sum_{i=2}^{k-1}\left\{ g_{k-1}\cdots g_{i+1}
%\left(\frac{(q^{\frac{1}{2}}-q^{-\frac{1}{2}})g_i+
%2}{2(q^{\frac{1}{2}}+q^{-\frac{1}{2}})}\right)
\left(\frac{dg_i}{dq}\right) g_{i-1}g_2g_1^2g_2\cdots g_{k-1}\right.\\ \\
&+&\left. g_{k-1}\cdots g_2g_1^2g_2\cdots g_{i-1}
%\left(\frac{(q^{\frac{1}{2}}-q^{-\frac{1}{2}})g_i
%+2}{2(q^{\frac{1}{2}}+q^{-\frac{1}{2}})}\right)
\left(\frac{dg_i}{dq}\right) g_{i+1}\cdots g_{k-1}\right\} \\
&+& g_{k-1}\cdots g_2\left(\frac{d}{dq}g_1^2\right) g_2\cdots g_{k-1}\\ \\
&{\buildrel q=1\over=}& \sum_{i=2}^{k-1}\left\{\frac{1}{2}s_{k-1}
\cdots s_{i+1}s_is_{i+1}\cdots s_{k-1}+\frac{1}{2}s_{k-1}\cdots s_{i+1}
s_is_{i+1}\cdots s_{k-1}\right\}\\ \\
&+ & s_{k-1}\cdots s_2s_1s_2\cdots s_{k-1}\\ \\
&=& \sum_{i=1}^{k-1}(i,k)=d_k.
\end{eqnarray*}
\qed
Finally, let us explain a connection between the multiplicative
Jucys--Murphy elements $D_k$ and the pure braid group $P_n$. More
precisely, let us consider the infinitesimal deformations of the pure
braid group $P_n$, and the multiplicative Jucys--Murphy elements $D_k$.
To start, it is convenient to remind a few definitions.
By definition, the pure braid group $P_n$ is a kernel of the natural
homomorphism
$$ B_n\to S_n, \ \ g_i\mapsto s_i=(i,i+1),
$$
where $s_i=(i,i+1)\in S_n$, $1\le i\le n-1$, denote the transposition
that interchanges $i$ and $i+1$, and fixes all other elements of $[1,n]$.
We have an exact sequence:
$$1\to P_n\to B_n\to S_n\to 1.
$$
It is well--known (see, e.g. [B]), that $P_n$ is generated by the
following elements
$$g_{ij}=(g_{j-1}g_{j-2}\cdots g_{i+1})g_i^2(g_{j-1}g_{j-2}\cdots
g_{i+1})^{-1}, \ 1\le i<j\le n,
$$
subject the following relations:
%where $1\le i<j\le n$, and $1\le k<m\le n$, and $j<m$:
\begin{eqnarray}
&& g_{ij}g_{kl}=g_{kl}g_{ij}, \ {\rm if~all} \ i,j,k,l \ {\rm
are~distinct,} \label{4.8}\\
&& g_{ij}g_{ik}g_{jk}=g_{ik}g_{jk}g_{ij}=g_{jk}g_{ij}g_{ik}, \ {\rm if} \
1\le i<j<k\le n,\label{4.9}\\
&& g_{ik}g_{jk}g_{jl}g_{ij}=g_{jk}g_{jl}g_{ij}g_{ik}, \ {\rm if} \ 1\le
i<j<k<l\le n. \label{4.10}
\end{eqnarray}
%\begin{equation}
%g_{ij}g_{km}g_{ij}^{-1}=\left\{\begin{array}{ll} g_{km},& k<i;\\ \\
%g_{jm}^{-1}g_{km}g_{jm}, & k=i;\\ \\
%g_{jm}^{-1}g_{im}^{-1}g_{jm}g_{im}g_{km}g_{im}^{-1}g_{jm}^{-1}
%g_{im}g_{jm}, & i<k<j;\\ \\
%g_{jm}^{-1}g_{im}^{-1}g_{km}g_{im}g_{jm}, & k=j;\\ \\
%g_{km} & k>j.
%\end{array}\right.
%\end{equation}
One can show using only the relations (\ref{4.8})--(\ref{4.10}) that the
elements
$$D_k=g_{1,k}g_{2,k}\cdots g_{k-1,k}\in P_n, \ 2\le k\le n,
$$
are mutually commute.
Now let us consider the infinitesimal deformation
$g_{ij}\mapsto 1+\epsilon X_{ij}$, of the pure braid group $P_n$. It is
easy to see that the coefficients of $\epsilon^2$ on
both sides of relations (\ref{4.8})--(\ref{4.10}) coincide with the
defining relations (\ref{3.1})--(\ref{3.2})
for the braid algebra $\B_n$, and the element $D_k$ transforms to the
Jucys--Murphy elements $d_k$ (more precisely, $D_k=1+\epsilon
d_k+o(\epsilon^2)$).
\section{Quadratic Algebra ${\cal G}_n$}
\label{sec:quad}
\neweq
In the recent paper of the author and S.~Fomin [FK] the
quadratic algebra $\E_n^t$ was introduced. This algebra is closely related
to the theory of
quantum Schubert polynomials [FK], and their multiparameter deformation
[K]. In this Section we introduce the quadratic algebra $\G_n$, which is a
further generation of the quadratic algebra $\E_n^t$.
\begin{de}\label{d1.1} Define the algebra $\G_n$ (of type $A_{n-1}$) as the
quadratic algebra (say, over $\Z$) with generators $[ij]$, $1\le i<j\le
n$, which satisfy the following relations
\begin{equation}
\label{1.1}
\begin{array}{rl}
&[ij][jk]=[jk][ik]+[ik][ij]\,,\\[.1in]
&[jk][ij]=[ik][jk]+[ij][ik],
\textrm{\ \ for } i<j<k\,;\quad\ \ \\[.1in]
\end{array}
\end{equation}
\begin{equation}
\label{1.2}
\begin{array}{rl}
&~[ij][kl]=[kl][ij],\\[.1in]
&~\textrm{whenever } \{i,j\}\cap\{k,l\}=\phi\,,
i<j, \textrm{ and }k<l \,.
\end{array}
\end{equation}
\end{de}
The quadratic algebra $\E_n^t$, [FK], is the quotient of the algebra
$\G_n$ by the two--side ideal generated by the relations
$$[ij]^2-t_{ij}=0,
$$
where the $t_{ij}$, for $1\le i<j\le n$, are a set of commuting
parameters.
From (2.1) follows that the generators $[ij]$ satisfy the
classical Yang--Baxter equation (CYBE)
\begin{equation}
[[ij],[ik]+[jk]]+[[ik],[jk]]=0, \ i<j<k, \label{2.2a}
\end{equation}
where the external brackets stand for the usual commutator: $[a,b]=ab-ba$.
\begin{theorem}\label{t1.1} The Hilbert series of the algebra $\G_n$ is
given by
\begin{equation}
\label{1.3}
H(\G_n;t)=\frac{1}{(1-t)(1-2t)\cdots (1-(n-1)t)}.
\end{equation}
\end{theorem}
Theorem~\ref{t1.1} is a corollary of the following result which gives a
description of an additive basis in the algebra $\G_n$:
\begin{theorem}\label{t1.2} A linear basis of the algebra $\G_n$ is given by
the following set of noncommutative monomials
\begin{equation}
\label{1.4}
Z_2\cdot Z_3\cdot\ldots\cdot Z_n,
\end{equation}
where
\begin{eqnarray*}
Z_2 & = & \{\phi ,[12]\} \\
& \cdots & \\
Z_k & = & \{\phi\}\cup\{ [i_1k]\cdots [i_lk]~\vert ~1\le i_j<k,~1\le
j\le l\} ,\ \ 2\le k\le n.
\end{eqnarray*}
\end{theorem}
A monomial in $[ij]$'s is called to be in {\it normal form}, if it
belongs to the set (\ref{1.4}).
A proof of Theorem~\ref{t1.1} is similar to a proof of
Corollary~\ref{c3.2} given by X.-S.~Lin, [L], Theorem~2.3. By
this reason we omit a proof.
\begin{rem}\label{r1.2} {\rm It is clear
that~~~~$H(Z_k;t)=\ds\frac{1}{1-(k-1)t}$, and }
\begin{equation}
H(\G_n;t)=\prod_{k=2}^nH(Z_k;t). \label{1.5}
\end{equation}
\end{rem}
\begin{cor}\label{c1.1} Let $\G_n'$ be a commutative version of the
algebra $\G_n$, i.e. let us assume that the relations (\ref{1.2}) are
valid for all $i,j,k,l$. Then
$$H(\G_n';t)=(1-t)^{-\left(\begin{array}{l}\small{n}\\\scriptsize{2}\end{array}\right)}.
$$
\end{cor}
\section{Dunkl elements}
\label{sec:dunkl}
\neweq
The Dunkl elements $\theta_j$, for $j=1,\ldots ,n$, in the quadratic
algebra $\G_n$ are defined by
\begin{equation}
\label{2.1}
\theta_j=-\sum_{1\le i<j}[ij]+\sum_{j<k\le n}[jk].
\end{equation}
For the quadratic algebra $\E_n^t$ the Dunkl elements (\ref{2.1})
coincide with those introduced in [FK], Section~5.
The following result is well--known (cf., e.g. [Ki], Theorem~1.4, or
[FK], Lemma~5.1)
\begin{lem}\label{l2.1} The classical Yang--Baxter equation (\ref{2.2a}),
together
with the commutativity relation (\ref{1.2}), imply that the Dunkl
elements $\theta_j$ commute pairwise.
\end{lem}
There exists an obvious relation between Dunkl's elements $\theta_1,\ldots
,\theta_n$ in the quadratic algebra $\G_n$, namely, $\theta_1+\cdots
+\theta_n=0$. The result below shows that in the algebra $\G_n$ the Dunkl
elements $\theta_1,\ldots ,\theta_{n-1}$ are algebraically independent
(cf. Theorem~\ref{t5.1}).
\begin{theorem}\label{t2.1} Let $\H_n$ be a commutative subalgebra of $\G_n$
generated by the Dunkl elements $\theta_j$, $j=1,\ldots ,n$. Then a map
$\theta_i\to x_i$, $1\le i\le n-1$, defines an isomorphism
$$\H_n\cong\Z [x_1,\ldots ,x_{n-1}],
$$
\end{theorem}
\begin{problem}\label{pr2.1} To describe the two--side ideals $\R\subset\G_n$
such that
\begin{equation}
H\left(\H_n/\R\cap\H_n;t\right)=[n]!:=\prod_{j=1}^n\frac{1-t^j}{1-t}.\label{2.2}
\end{equation}
\end{problem}
The examples of two--side ideals in the algebra $\G_n$ with the property
(\ref{2.2}) will be given in Sections~7 and 8. These examples are related
to the quantum cohomology ring of the flag variety [FK], and the
multiparameter deformation of Schubert polynomials [K].
\begin{ex}\label{e2.1} {\rm Let us take $n=3$. In the algebra $\G_3$ we have
the following relations
\begin{eqnarray*}
&&\theta_1+\theta_2+\theta_3=0,\\
&&\theta_1\theta_2+\theta_1\theta_3+\theta_2\theta_3+[12]^2+[13]^2+[23]^2=0,\\
&&\theta_1\theta_2\theta_3+[12]^2\theta_3+\theta_1[23]^2-[13]^2\theta_1-
\theta_3[13]^2=0, \\
&&[\theta_2, [23]^2]=[\theta_1,[13]^2],
\end{eqnarray*}
where $[a,b]=ab-ba$ is the usual commutator.}
\end{ex}
\section{Dunkl and Jucys--Murphy elements}
\label{sec:djm}
\neweq
Let us start with the definition of the extended affine braid group
$\wt B_n$.
\begin{de}\label{d6'.1} The extended affine braid group $\wt B_n$ is
a group with generators
$$g_0, g_1,\ldots ,g_{n-1}, w
$$
which satisfy the following relations
\begin{eqnarray}
&& g_ig_j=g_jg_i, \ \ |i-j|\ge 2, \ 0\le i,j\le n-1, \label{6'.1}\\
&& g_ig_{i+1}g_i=g_{i+1}g_ig_{i+1}, \ \ 0\le i\le n-1, \label{6'.2}\\
&& wg_i=g_{i-1}w,\ \ 0\le i\le n-1, \label{6'.3}
\end{eqnarray}
with indices understood as elements of $\Z /n\Z$.
\end{de}
It follows from (\ref{6'.3}) that $w^n$ is a central element.
There exists a canonical homomorphism $\pi$ from the extended affine
braid group $\wt B_n$ to the classical braid group $B_n$. On the
generators $g_i$ and $w$ the homomorphism $\pi$ is given by the
following rules
\begin{equation}
\pi (g_i)=g_i, \ 1\le i\le n-1, \label{6'.4}
\end{equation}
$$\pi (w)=g_{n-1}g_{n-2}\cdots g_2g_1.
$$
It follows from (\ref{6'.4}) that
$$\pi (g_0)=\pi (w)g_1\pi (w)^{-1}=g_{n-1}g_{n-2}\cdots
g_2g_1g_2^{-1}\cdots g_{n-2}^{-1}g_{n-1}^{-1}.
$$
Now we are going to define the Dunkl
ele\-ments $Y_1,\ldots ,Y_n$, and the dual Dunkl elements $Y_1^*,\ldots
,Y_n^*$ in the extended affine braid group.
\begin{de}\label{d6'.2} For each $i=1,\ldots ,n$,
we define (cf. [KN]; [Ch1]) the Dunkl and dual Dunkl elements $Y_i$ and
$Y_i^*$ respectively, by the following formulae
\begin{eqnarray}
Y_i& = & g_ig_{i+1}\ldots g_{n-1}wg_1^{-1}\ldots g_{i-1}^{-1},
\label{6'.5}\\
Y_i^*& = &g_i^{-1}g_{i+1}^{-1}\ldots g_{n-1}^{-1}wg_1\ldots g_{i-1}.
\label{6'.6}
\end{eqnarray}
\end{de}
Note that $Y_1=g_1\ldots g_{n-1}w$, $Y_n=wg_1^{-1}\ldots g_{n-1}^{-1}$,
$Y_1^*=wg_1\ldots g_{n-1}$ and\break $Y_n^*=g_1^{-1}\ldots g_{n-1}^{-1}w$.
\vskip 0.3cm
\hskip -0.6cm The Dunkl elements satisfy the following commutation relations with
$g_1,\ldots ,g_{n-1}$:
\begin{eqnarray}
g_iY_{i+1}g_i &=& Y_i,~~~~~g_iY_j = Y_jg_i \ \ (j\ne i,i+1),
\label{6'.7}\\
g_iY_i^*g_i &=& Y_{i+1}^*, \ \ g_iY_j^* = Y_jg_i \ \ (j\ne i,i+1).\label{6'.8}
\end{eqnarray}
It is clear from our construction, that the image in the braid group
$B_n$ of the dual Dunkl element $Y_k^*$, $2\le k\le n$, under the
canonical homomorphism $\pi :\wt B_n\to B_n$ coincides with the
multiplicative Jucys--Murphy element $D_k$, namely
\begin{eqnarray*}
\pi (Y_k^*) &= & g_k^{-1}g_{k+1}^{-1}\cdots g_{n-1}^{-1}\pi (w)g_1\cdots
g_{k-1}\\
& = & g_{k-1}\cdots g_2g_1^2g_2\cdots g_{k-1}{\buildrel (\ref{4.6})\over
=}D_k, \ \ 2\le k\le n,
\end{eqnarray*}
and $\pi (Y_1^*)=1$.
\begin{lem}\label{l6'.1} The Dunkl elements $Y_k$ (resp. the dual Dunkl
elements $Y_k^*$) commute pairwise.
\end{lem}
Let us illustrate the main idea of the proof of Lemma~\ref{l6'.1} on some
example. Let us take $n=5$ and prove that $Y_2^*Y_3^*=Y_3^*Y_2^*$. Indeed,
\begin{eqnarray*}
Y_2^*Y_3^* &=& g_2^{-1}g_3^{-1}g_4^{-1}wg_1g_3^{-1}g_4^{-1}wg_1g_2
=g_2^{-1}g_3^{-1}g_4^{-1}g_0g_2^{-1}g_3^{-1}g_4g_0w^2\\
& = & g_2^{-1}g_3^{-1}g_2^{-1}g_4g_3^{-1}g_0g_4g_0w^2
=g_3^{-1}g_2^{-1}g_3^{-1}g_4g_3^{-1}g_4g_0g_4w^2.
\end{eqnarray*}
Similarly,
$$Y_3^*Y_2^*=g_3^{-1}g_4^{-1}wg_1g_3^{-1}g_4^{-1}wg_1
=g_3^{-1}g_2^{-1}g_4^{-1}g_3^{-1}g_0g_4w^2.
$$
It is easy to see that in $\wt B_5$ we have
$$g_3^{-1}g_2^{-1}g_3^{-1}g_4g_3^{-1}g_4g_0g_4
=g_3^{-1}g_2^{-1}g_4^{-1}g_3^{-1}g_0g_4.
$$
\vskip 0.3cm
Now let us study the quasi--classical limit of the Dunkl
element $Y_k$. The first step is to consider $Y_k$ as an element of the
extended affine Hecke algebra $H(\wt W)$, [KN], Section~2; [Ch1], [Ch2].
In other
words, we have to add to the relations (\ref{6'.1})--(\ref{6'.3}) the new
ones
\begin{equation}
g_i^2=(t-1)g_i+t, \ \ 0\le i\le n\label{6'.7'}
\end{equation}
where $t$ is a new parameter. The next step is to consider a
representation of the extended affine Hecke algebra $H(\wt W)$ in the
algebra $\D_{q,x}[W]$ of $q$--difference operators with permutations,
see, for example, [Ch1], [KN]. Follow [KN], we define the elements
$T_i$, $i=0,1,\ldots ,n-1$, in $\D_{q,x}[W]$ by
\begin{eqnarray*}
T_i&=&t+\frac{x_{i+1}-tx_i}{x_{i+1}-x_i}(s_i-1),\ \ i=1,\ldots ,n-1,\\
T_0&=&t+\frac{x_1-tqx_n}{x_1-qx_n}(s_0-1).
\end{eqnarray*}
Here $s_1,\ldots,s_{n-1}$ are the standard generators of the symmetric
group $S_n$, and\break $s_0=ws_1w^{-1}$, where $w=s_{n-1}s_{n-2}\cdots s_1\tau_1$,
and $\tau_1=T_{q,x_1}$ is the $q$--shift operator:
$$(\tau_1f)(x_1,\ldots ,x_n)=f(qx_1,x_2,\ldots ,x_n).
$$
%it is clear that
%\begin{eqnarray*}
%(w^{-1}f)(x_1,\ldots ,x_n)&=&f(q_1x_n,x_1,x_2,\ldots ,x_{n-1}),\\
%(wf)(x_1,\ldots ,x_n)&=&f(x_2,x_3,\ldots , x_n,q^{-1}x_1),\\
%s_0f)(x_1,\ldots ,x_n)&=&f(qx_n,x_2,\ldots ,x_{n-1},q^{-1}x_1).
%\end{eqnarray*}
One can check that the elements $T_i$, $0\le i\le n-1$, and $w$, satisfy the
relations (\ref{6'.1})--(\ref{6'.3}) and (\ref{6'.7'}), and define
a representation of the extended affine Hecke algebra $H(\wt W)$ in the
algebra $\D_{q,x}[W]$ of $q$--difference operators with permutations.
In this representation, the Dunkl element $Y_k$ (resp. $Y_k^*$),
$1\le k\le n$, up to a power of $t$ coincides (see [Ch1], [Ch2], [KN]) with
the Dunkl--Cherednik operator ${\bf Y}_k$:
\begin{eqnarray}
{\bf Y}_k&=&t^{-n+2k-1}T_kT_{k+1}\ldots T_{n-1}wt_1^{-1}\ldots
T_{k-1}^{-1}, \label{6'.8'}\\
{\bf Y}_k^*&=&t^{n-2k+1}T_k^{-1}T_{k+1}^{-1}\ldots T_{n-1}^{-1}w
T_1\ldots T_{k-1}. \label{6'.8"}
\end{eqnarray}
In order to understand better the quasi--classical behavior of the
Dunkl--Cherednik operator ${\bf Y}_k$, it is convenient to rewrite
slightly the formulae (\ref{6'.8'})--(\ref{6'.8"}). Namely, let us define
($1\le i<j\le n$) the following operators acting on the ring of
polynomials $\Z [t,t^{-1}][x_1,\ldots ,x_n]$
$$\overline T_{ij}=1+\frac{(1-t^{-1})x_j}{x_i-x_j}(1-s_{ij})=
t^{-1}T_{ij}s_{ij},
$$
where $s_{ij}$ is the exchange operator for variables $x_i$, $x_j$. Then
\begin{equation}
{\bf Y}_i=\overline T_{i,i+1}\overline T_{i,i+2}\ldots
\overline T_{i,n}\tau_i\overline T_{1,i}^{-1}\ldots
\overline T_{i-1,i}^{-1}, \label{6'.9'}
\end{equation}
where $\tau_i=s_is_{i+1}\ldots s_{n-1}ws_1\ldots s_{i-1}$ is the
$q$--shift operator: $\tau_i=\tau_{x_i}=T_{q,x_i}$.
Finally, in the quasi--classical limit $q\to 1$ with
rescaling $t=q^{\beta}$, we obtain from (\ref{6'.9'}) that
\begin{equation}
\D_j:=\lim_{q\to 1}~\frac{1-{\bf Y}_j}{1-q}=
x_j\frac{\partial}{\partial x_j}+\beta\sum_{i<j}x_j\partial_{ij}-
\beta\sum_{k>j}\partial_{jk}x_k. \label{6'.9"}
\end{equation}
Note that these Dunkl operators $\D_k$ commute with each other, i.e.
$[\D_i,\D_j]=0$.
More generally, let us define the affine extension of the algebra
$\G_n$:
\begin{de} Define the algebra $\wt\G_n$ as the algebra (say over
$\Z [q,q^{-1}]$)with generators
$$[ij], \ \ 0\le i<j\le n;\ \ x_i,\ 1\le i\le n, \ \ {\rm and} \ w,
$$
which satisfy the relations (\ref{1.1}) and (\ref{1.2}) for $i,j,k,l\in
[1,n]$, and the
following additional relations
\begin{eqnarray}
x_ix_j&=&x_jx_i, \ \ 1\le i,j\le n,\label{6'.9}\\
x_i[ab]&=&[ab]x_i, \ {\rm if} \ i\ne a,b, \label{6'.10}\\
x_j[ij]&=&[ij]x_i+1,\ \ x_i[ij]=[ij]x_j-1\ \ 1\le i<j\le n, \label{6'.11}\\
wx_i&=&q^{\delta_{1,i}}x_{i-1}w, \label{6'.11'}\\
w[ij]&=&q^{-\delta_{i,0}}[i-1,j-1]w, \label{6'.11''}
\end{eqnarray}
with indices understood as elements of $\Z /n\Z$.
\end{de}
It is clear from (\ref{6'.11}), that
\begin{equation}
[x_i+x_j,[ij]]=0. \label{6'.12}
\end{equation}
We define the Dunkl elements $\wt\theta_j$ in the algebra $\wt\G_n$ by the
following rule
\begin{equation}
\wt\theta_j=x_j+\theta_j, \ \ 1\le j\le n. \label{6'.13}
\end{equation}
The following result is well-known, see, e.g. [Ki]:
\begin{lem}\label{l6'.2} The classical Yang--Baxter equation (\ref{2.2a}),
together with commutativity relation (\ref{1.2}) and relation
(\ref{6'.12}), imply that the Dunkl elements $\wt\theta_j=x_j+\theta_j$
commute pairwise.
\end{lem}
{\it Proof.} If $i<j$, then
$$[\wt\theta_i,\wt\theta_j]=[x_i,x_j]+[x_i,\theta_j]+[\theta_i,x_j]+
[\theta_i,\theta_j]=[x_i+x_j,[ij]]=0.
$$
\qed
Let us introduce the algebra $\wt\G_n^0$ which is the quotient of the
algebra $\wt\G_n$ by the two--side ideal generated by the relation
\begin{equation}
[ij]^2=0, \ \ 1\le i<j\le n. \label{6'.14}
\end{equation}
\begin{de}\label{d6'.3} For $1\le i<j\le n$ let us define an element
$T_{ij}\in\wt\G_n^0$ by the following formula (cf. [KN])
\begin{equation}
T_{ij}=t-(x_j-tx_i)[ij]. \label{6'.15}
\end{equation}
\end{de}
It follows from (\ref{6'.11}) and (\ref{6'.14}), that
$$T_{ij}^2=(t-1)T_{ij}+t.
$$
Our next goal is to check that the elements
$\wt T_i:=T_{i,i+1}$, $1\le i\le n-1$, satisfy the Coxeter relations.
\begin{pr}\label{p6'.1} We have the following relation in the algebra
$\wt\G_n^0$:
\begin{equation}
T_{ab}T_{bc}T_{ab}=T_{bc}T_{ab}T_{bc}, \ \ a<b<c. \label{6'.16}
\end{equation}
\end{pr}
{\it Proof.} Direct computation based on (\ref{6'.11}) and the following
relation in the algebra $\wt\G_n^0$:
$$[ab][bc][ab]=[bc][ab][bc], \ \ 1\le a<b<c\le n.
$$
\qed
As a corollary, we see that the elements $w$, $\wt T_0=w\wt T_1w^{-1}$,
$\wt T_1,\ldots ,\wt T_{n-1}$ generate
the representation of the extended affine Hecke algebra $H(\wt W)$. In
this representation the quasi--classical limit $q\to 1$ with rescaling
$t=q^{\beta}$ of the Dunkl element $t^{-n+2j-1}Y_j$ is equal to
\begin{equation}
x_j\frac{\partial}{\partial x_j}+\beta\sum_{i<j}x_j[ij]-
\beta\sum_{k>j}[jk]x_k. \label{6'.17}
\end{equation}
The element (\ref{6'.17}) does not belong to the algebra $\wt\G_n$, but
its further extension $\wt{\wt\G}_n$ which is an analog of the double
affine Hecke algebra, introduced by I.~Cherednik. It is an interesting
problem to understand some connections between the Schubert calculus and
the Dunkl--Cherednik operators.
\begin{con} {\rm (Nonnegativity conjecture, cf. [FK], Conjecture~8.1)}
For any $w\in S_n$, the Schubert polynomial ${\s}_w$ evaluated at the Dunkl
elements $\wt\theta_1,\ldots ,\wt\theta_n$ belongs to the positive cone
$\wt\G_n^+$, where $\wt\G_n^+$ is the cone of all nonnegative integer
linear combinations of all (noncommutative in general) monomials in the
generators $x_i$, $1\le i\le n$, and $[ij]$, $1\le i<j\le n$, of the
algebra $\wt\G_n^0$.
\end{con}
\newpage
\section{Quadratic algebra $\E_n^t$}
\label{sec:ent}
\neweq
In this and the next sections we study some interesting quotients of the the
quadratic algebra $\G_n$.
\begin{de}\label{d5.1} {\rm (cf. [FK], Section~15)} Define the algebra
$\E_n^t$ as
the algebra over $\Z$ with generators $[ij]$, $1\le i<j\le n$, which
satisfy the relations (\ref{1.1}), (\ref{1.2}) and the additional relations
\begin{equation}
([ij],[kl]^2)=0, ~~for~ all~~i,j,k,l.
\label{5.1}
\end{equation}
\end{de}
The relations (\ref{5.1}) mean that the squares $[kl]^2$ belong to the
center of $\E_n^t$. Let us put $t_{ij}=[ij]^2$, and consider the ring of
polynomials $\Z [t]:=\Z [~t_{ij},~1\le i<j\le n]$.
\begin{con}\label{c5.1} {\rm ([FK])} The algebra
$\E_n^t$ is a finite dimensional module over the ring of polynomials $\Z
[t]$.
\end{con}
\begin{theorem}\label{t5.1} {\rm ([FK], Conjecture~15.1; [P])} Let
$\H_n^t$ be a commutative subalgebra of $\E_n^t$ generated by the Dunkl
elements $\theta_1,\ldots ,\theta_n$. Then
$$\H_n^t\cong\Z [t][\theta_1,\ldots ,\theta_n]/J,
$$
where the ideal $J$ is generated by the generalized elementary functions
\begin{equation}
e_m(X_n|t):=\sum_l\sum_{\begin{array}{c} \scriptstyle{1\le i_1<\cdots
<i_l\le n}\\ \scriptstyle{j_1>i_1,\ldots ,j_l>i_l}\end{array}}
e_{m-2l}(X_{\overline{I\cap J}})\prod_{k=1}^lt_{i_kj_k},
\label{5.2}
\end{equation}
where $i_1,\ldots ,i_l,j_1,\ldots ,j_l$ should be distinct elements of the
set $\{ 1,\ldots ,n\}$, and $X_{\overline{I\cap J}}$ denotes set of
variables $x_a$, for which the subscript $a$ is neither one of the $I_k$
nor one of the $J_k$.
\end{theorem}
One can check that the relations (\ref{5.1}) are
equivalent to the following ones
$$(\theta_i,[kl]^2)=0, \ \ {\rm for~ all} \ \ i,k,l.
$$
The relations (\ref{5.1}) in the algebra $\E_n^t$ imply some nontrivial
relations between monomials in the space
$Z:=Z_2\cdot Z_3\cdot\ldots\cdot Z_n$. We will
describe part of these relations in Proposition~\ref{p5.1} below, but before
that, let us introduce some additional notations. For given $n\ge 3$ and
$k$, $1\le k<n$, let us denote by $\Lambda_{n,k}$ a set of all sequences
of integer numbers $A=(a_1,\ldots ,a_k)$, such that $a_1+\cdots +a_k=n-2$,
and $a_i\in\Z_{>0}$. For each sequence $A$ from the set $\Lambda_{n,k}$
let us define a monomial $[A]$ in the space $Z$ as the following ordered
product
$$[A]=\prod_{j=1}^k\prod_{l=1}^{a_j}[n-k+l-2-\sum_{s=1}^j(a_s-1),
~n-k+j-1].
$$
For example, for $n=7$ and $A=(2,1,2)$ we have
$$[A]=[24][34][25][16][26].
$$
\begin{pr}\label{p5.1}
\begin{equation}
F_n:=\sum_{i=1}^{n-1}[i,n][i+1,n]\cdots [n-1,n][1,n][2,n]\cdots [i,n]
\label{5.3}
\end{equation}
$$=\sum_{k=2}^{n-1}(-1)^{n-k-1}(t_{kn}-t_{k-1,n})
\sum_{A\in\Lambda_{n,n-k}}[A].~~~~~
$$
\end{pr}
{\it Proof.} By induction, using the following
formula
\begin{eqnarray*}
&& \Phi_{n+1}\cdot [12]-[12]\cdot\Phi_{n+1}=F_{n+1},\ \ {\rm where}\\
&&\Phi_n:=[2,n]\cdots [n-1,n][2,n]+\cdots +[n-1,n][2,n]\cdots
[n-2,n][n-1,n].
\end{eqnarray*}
\qed
Let us make the following simple but useful observation. The action of
the symmetric group, namely, $w([i_1j_1]\cdots [i_kj_k])=[w(i_1)w(j_1)]
\cdots [w(i_k)w(j_k)]$, $w\in S_n$, transforms every specific identity in
$\G_n$ (resp. $\E_n^t$, $\E_n^0$) into an orbit of identities. As an
example, the identity (in $\E_n^t$)
$$[12][13][14][12]+[13][14][12][13]+[14][12][13][14]=
(t_{34}-t_{24})[13][23]-(t_{24}-t_{14})[12][13]
$$
(a special case of Proposition~\ref{p5.1}) immediately implies a more general
identity
\begin{eqnarray*}
&&[ab][ac][ad][ab]+[ac][ad][ab][ac]+[ad][ab][ac][ad]\\
&&=([cd]^2-[bd]^2)[ac][bc]-([bd]^2-[ad]^2)[ab][ac].
\end{eqnarray*}
Proposition below describes some interesting relations in the quadratic
algebra $\E_n^t$. They can be interpreted as the Pieri rules for the
classical and quantum Schubert polynomials, [FK], [P], and for the
multiparameter deformation of Schubert polynomials [K], [P].
\begin{pr}\label{p5.2} {\rm ([P])} Let $A$ be a subset in $[1,\ldots ,n]$,
and let us put $\theta_A=\ds\prod_{i\in A}\theta_i$. Then in the algebra
$\E_n^t$ we have the following relation
$$e_m(\theta_A|t)=\sum [i_1j_1]\cdots [i_mj_m],
$$
where the summation is taken over all sequences $I=\{ i_1,\ldots ,i_m\}$,
$J=\{ j_1,\ldots ,j_m\}$ such that
i)~~~ $I\subset A$, \ $J\subset [1,n]\setminus A$,
ii)~~ $i_1,\ldots ,i_m$ are all distinct,
iii)~ $1\le j_1\le j_2\le\cdots\le j_m\le n$.
\end{pr}
\begin{problem} Find a nontrivial faithful representation of the
quadratic algebra $\E_n^t$.
\end{problem}
\section{Quadratic algebra $\E_n^0$ }
\label{sec: En0}
\neweq
In this Section we study the properties of the algebra $\E_n^t$ when all
parameters $t$ are equal to zero. Hence, in this Section we assume that
\begin{equation}
[ij]^2=0, \ \ 1\le i<j\le n.
\label{6.1}
\end{equation}
This algebra was introduced and studied at first in [FK]. It follows
from Theorem~\ref{t1.2} that any element of $\E_n^0$ can be presented as
$\Z$-linear combination of monomials $z$ of the following form
\begin{equation}
z=z_1z_2\cdots z_n,
\label{6.2}
\end{equation}
where $z_i\in Z_i$, and ($2\le k\le n$)
\begin{equation}
Z_k:=Z_{n,k}=\{\phi\}\cup\{ [i_1k]\cdots [i_lk]~|~1\le i_l<k, \ 1\le
j\le l\}. \label{6.3}
\end{equation}
In the algebra $\G_n$ the monomials in the set $Z_k$ are linearly
independent (Theorem~\ref{t1.2}). However, in the quotient algebra $\E_n^0$
%(or $\E_n^t$)
some new relations between the monomials in $Z_k$ are appeared. The
Proposition~\ref{p6.1} below is a special case of Proposition~\ref{p5.1} and
gives the relations in subalgebra $Z_m\subset\E_n^0$, but not only
in the space $Z=Z_2\cdot Z_3\cdot\ldots\cdot Z_n$.
\begin{pr}\label{p6.1} {\rm (cf. [FK])} Let $a_1,\ldots ,a_m$ be a sequence
of pairwise distinct integer numbers, $1\le a_i<n$, $m\le n$. Then
\begin{equation}
a_1a_2\cdots a_ma_1+a_2a_3\cdots a_1a_2+\cdots +a_ma_1a_2\cdots
a_{m-1}a_m=0.
\label{6.4}
\end{equation}
Here we used an abbreviation $a_j=[a_jn]$.
\end{pr}
{\it Proof.} It is enough to prove that
\begin{equation}
F_n:=\sum_{i=1}^{n-1}[i,n][i+1,n]\cdots [n-1,n][1,n][2,n]\cdots [i,n]=0.
%12\cdots n-11+23\cdots n-112+\cdots +n-112\cdots n-2n-1=0.
\label{6.5a}
\end{equation}
Let us prove (\ref{6.5a}) by induction. Thus we may assume that $F_k=0$
for $k\le n$. Let us consider the following element in $Z_{n+1}$:
$$\Phi_n=[2,n+1]\cdots [n,n+1][2,n+1]+\cdots +[n,n+1][2,n+1]\cdots
[n-1,n+1][n,n+1].
$$
By induction assumption, we have $\Phi_n=0$. On the other hand, it is
easy to check that
\begin{equation}
\Phi_n\cdot [12]-[12]\cdot\Phi_n=F_{n+1}. \label{6.5b}
\end{equation}
The last equality shows that $F_{n+1}=0$.
\qed
\begin{ex} {\rm If $m=3$, the relation (\ref{6.4}) has the following form
($a<b<n$)
\begin{equation}
aba+bab=0, \ \ {\rm in~the~algebra}\ \ \E_n^0
\label{6.5}
\end{equation}
Similarly, if $a,b,c$ are distinct, $<n$, then}
\begin{equation}
abca+bcab+cabc=0, \ \ {\rm in~the~algebra}\ \ \E_n^0
\label{6.6}
\end{equation}
\end{ex}
\vskip 0.3cm
We don't know how to describe all
relations between monomials from the space $Z_k$ for fixed $k$, but
we can show that if
$k\le 5$, then all relations between monomials in $Z_k$ follows from
(\ref{6.4}). This observation allows us to compute the dimensions of the
homogeneous components $\E_{n,k}^0$ for $1\le k\le 5$, and also
$\dim\E_{n,6}^0$. We summarize the
results in the following
\begin{pr}\label{p6.2}~
\vskip 0.3cm
$\bullet$ ${\rm dim}\,\E_{n,1}^0=\left(\begin{array}{c}n\\
2\end{array}\right)$,
\vskip 0.3cm
$\bullet$ ${\rm dim}\,\E_{n,2}^0=3\left(\begin{array}{c}n\\
4\end{array}\right) +4\left(\begin{array}{c}n\\ 3\end{array}\right)$,
\vskip 0.3cm
$\bullet$ ${\rm dim}\,\E_{n,3}^0=15\left(\begin{array}{c}n\\
6\end{array}\right) +40\left(\begin{array}{c}n\\ 5\end{array}\right) +
30\left(\begin{array}{c}n\\ 4\end{array}\right) +3\left(\begin{array}{c}n\\
3\end{array}\right)$,
\vskip 0.3cm
$\bullet$ ${\rm dim}\,\E_{n,4}^0=105\left(\begin{array}{c}n\\
8\end{array}\right) +420\left(\begin{array}{c}n\\ 7\end{array}\right) +
610\left(\begin{array}{c}n\\ 6\end{array}\right) +366\left(\begin{array}{c}n\\
5\end{array}\right)$
\vskip 0.3cm
\hskip 2cm $+67\left(\begin{array}{c}n\\
4\end{array}\right) +\left(\begin{array}{c}n\\
3\end{array}\right)$,
\vskip 0.3cm
$\bullet$ ${\rm dim}\,\E_{n,5}^0=945\left(\begin{array}{c}n\\
10\end{array}\right) +5040\left(\begin{array}{c}n\\ 9\end{array}\right) +
10780\left(\begin{array}{c}n\\ 8\end{array}\right) +
11571\left(\begin{array}{c}n\\ 7\end{array}\right)$
\vskip 0.3cm
\hskip 2cm $+6285\left(\begin{array}{c}n\\ 6\end{array}\right) +
1480\left(\begin{array}{c}n\\ 5\end{array}\right) +
96\left(\begin{array}{c}n\\ 4\end{array}\right)$,
\vskip 0.3cm
$\bullet$ ${\rm dim}\,\E_{n,6}^0=10395\left(\begin{array}{c}n\\
12\end{array}\right) +69300\left(\begin{array}{c}n\\ 11\end{array}\right) +
195300\left(\begin{array}{c}n\\ 10\end{array}\right) +
299908\left(\begin{array}{c}n\\ 9\end{array}\right)$
\vskip 0.3cm
\hskip 2cm $+268674\left(\begin{array}{c}n\\ 8\end{array}\right) +
138545\left(\begin{array}{c}n\\ 7\end{array}\right) +
37456\left(\begin{array}{c}n\\ 6\end{array}\right) +
4231\left(\begin{array}{c}n\\ 5\end{array}\right)$
\vskip 0.3cm
\hskip 2cm $+106\left(\begin{array}{c}n\\ 4\end{array}\right)$.
\end{pr}
\vskip 0.3cm
\begin{con}\label{c6.1}
\begin{equation}
{\rm dim}\,\E_{n,k}^0=(2k-1)!!\left(\begin{array}{c}n\\ 2k\end{array}\right) +
(2k-1)!!\frac{4(k-1)}{3}\left(\begin{array}{c}n\\ 2k-1\end{array}\right)
\label{6.7}
\end{equation}
$$~~~~~~~~~~~~~~+(2k-3)!!\frac{(k-1)(k-2)(16k-3)}{9}
\left(\begin{array}{c}n\\ 2k-2\end{array}\right) +\cdots .
$$
\end{con}
{\it Proof of Proposition~\ref{p6.2}}. First of all, let us compute ${\rm
dim}\,Z_{n,k}$ for small values of $k$. It is clear that ${\rm
dim}\,Z_{n,1}=n-1$ and ${\rm dim}\,Z_{n,2}=(n-1)(n-2)$, because there are
no nontrivial relations of degree $<3$ in $Z_n$. In degree 3 we have to
take into account the relations (\ref{6.5}). Thus,
\vskip 0.3cm
${\rm dim}\, Z_{n,3}=(n-2){\rm dim}\, Z_{n,2}-\ds\left(\begin{array}{c}n-1\\
2\end{array}\right) =\frac{(n-1)(n-2)(2n-5)}{2}.$
\vskip 0.3cm
Similarly, in degree 4 we have to consider both the relations (\ref{6.5})
and (\ref{6.6}). Thus,
${\rm dim}\, Z_{n,4}=(n-2){\rm dim}\, Z_{n,3}-(n-2)\left(\begin{array}{c}n-1\\
2\end{array}\right) -2\left(\begin{array}{c}n-1\\
3\end{array}\right) $
\vskip 0.3cm
\hskip 1.6cm$=\ds\frac{(n-1)(n-2)(n-3)(3n-7)}{3}=24\left(\begin{array}{c}
n-1\\ 4\end{array}\right) +10\left(\begin{array}{c} n-1\\
3\end{array}\right) .$
\vskip 0.3cm
Using the same method, we can compute
\vskip 0.3cm
${\rm dim}\, Z_{n,5}=120\left(\begin{array}{c}n-1\\ 5\end{array}\right) +
86\left(\begin{array}{c}n-1\\ 4\end{array}\right) +
9\left(\begin{array}{c}n-1\\ 3\end{array}\right) . $
\vskip 0.3cm
When $k\ge 6$, some additional to (\ref{6.4}) relations between monomials
in $Z_k$ will appear. We describe such new relations in degree 6.
\begin{lem}\label{l6.3} Let $a,b,c,d$ be a set of distinct integer
numbers. Then
\begin{eqnarray}
&& abcdca+bcdcab+cabadc+dcabad+cdcaba \label{8.9"}\\
&& =abacdc+acdcba+bacdcb+cdabac+dabacd.\nonumber
\end{eqnarray}
where $a:=[an]$, \ $n\ge 5$.
\end{lem}
{\it Proof of Lemma~\ref{l6.3}.} Let us consider the difference
\begin{equation}
[ac]F_4(a,d,c,b)-F_4(a,b,c,d)[ac], \label{8.9'}
\end{equation}
where
\begin{eqnarray*}
F_4(a,b,c,d)&=&[an][bn][cn][dn][an]+[bn][cn][dn][an][bn]\\
&+&[cn][dn][an][bn][cn]+[dn][an][bn][cn][dn].
\end{eqnarray*}
First of all, the difference (\ref{8.9'}) is equal to zero. On the other
hand, using the relations in the algebra $\E_n^0$, one can transform the
expression (\ref{8.9'}) to the difference between the left and
the right hand sides of (\ref{8.9"}).
\qed
Using Lemma~\ref{l6.3}, we can compute
\vskip 0.4cm
${\rm dim}\, Z_{n,6}=720\left(\begin{array}{c}n-1\\ 6\end{array}\right) +
756\left(\begin{array}{c}n-1\\ 5\end{array}\right) +
187\left(\begin{array}{c}n-1\\ 4\end{array}\right) +
6\left(\begin{array}{c}n-1\\ 3\end{array}\right) .$
\vskip 0.2cm
Finally, in order to compute the dimensions of homogeneous components
$\E_{n,k}^0$, $1\le k\le 6$, we use an observation that monomials from
the set $Z_2\cdot Z_3\cdot Z_4\cdot Z_5\cdot Z_6$ form a basis in
$\displaystyle\bigoplus_{k=0}^6\E_{n,k}^0$.
The proof of Proposition~\ref{p6.2} is
finished.
\qed
\begin{con}\label{c6.2} {\rm (Factorization property)}
\begin{equation}
H(\E_n^0;t)=H(Z_{n,2};t)H(Z_{n,3};t)\cdots H(Z_{n,n};t). \label{6.8}
\end{equation}
\end{con}
\begin{ex}\label{e6.1} {\rm If $n=2$, then $Z_2=\{\phi\}\cup\{ 12\}$, and
$H(Z_2;t)=1+t$.
\vskip 0.3cm
If $n=3$, then
$Z_3=\{\phi\}\cup\{ 13, \ 23, \ 13\cdot 23,\ 23\cdot 13, \
13\cdot 23\cdot 13\},$ and
$$H(Z_3;t)=1+2t+2t^2+t^3.
$$
Now, let us consider $n=4$. Then ($a:=[an]$)
\begin{eqnarray*}
Z_{4,1}& = & \{ 1, 2, 3\},\\
Z_{4,2}& = & \{ 12, 13, 21, 31, 23, 32\},\\
Z_{4,3}& = & \{ 121, 123, 131, 132, 213, 312, 231, 232, 321\},\\
Z_{4,4}& = & \{ 1213, 1231, 1232, 1312, 1321, 2131, 2132, 3121, 2312, 2321\},\\
Z_{4,5}& = & \{ 12131, 12132, 12312, 12321, 13121, 21312, 21321, 31213,
23121\},\\
Z_{4,6}& = & \{ 121312, 121321, 123121, 131213, 213121, 231213\},\\
Z_{4,7}& = & \{ 1213121, 1231213, 2131213\},\\
Z_{4,8}& = & \{ 12131213\}.
\end{eqnarray*}
Hence, $H(Z_4;t)=(1+t)^2(1+t+t^2)(1+t^2)^2$, and
$$H(\E_4^0;t)=(1+t)^4(1+t^2)^2(1+t+t^2)^2.
$$
Formula for $H(\E_4^0;t)$ was appeared in [FK] and confirmed by J.-E.~Ross
[Ro], who also computed the Hilbert polynomial $H(\E_5^0;t)$.}
\end{ex}
\section{Quadratic algebra $\B_n^t$ }
\label{sec: Bnt}
\neweq
In this Section we define a certain quotient $\B_n^t$ of the braid
algebra $\B_n$, and formulate a conjecture that the algebras $\B_n^t$ and
$\E_n^t$ have the same Hilbert polynomials.
\begin{de}\label{d10.1} define the algebra $\B_n^t$ as the algebra over
$\Q$ with generators $X_{ij}$, $1\le i<j\le n$, which satisfy the
relations (\ref{3.1}), (\ref{3.2}) and the additional relations
\begin{equation}
[ij]^2=t_{ij}, \label{10.1}
\end{equation}
where the $t_{ij}$ for $1\le i<j\le n$, are a set of commuting parameters.
\end{de}
In the case when all parameters $t_{ij}$ are equal to zero, we denote the
algebra $\B_n^t$ by $\B_n^0$.
\begin{con}\label{c10.1} The algebra $\B_n^t$ is a finite dimensional
module over the ring of polynomials $\Q [t]:=\Q [t_{ij}, 1\le i<j\le n]$.
\end{con}
\begin{problem} To find all relations among the Jucys--Murphy
elements $d_j$,\break $2\le j\le n$, and find the Hilbert polynomial of
commutative subalgebra $\K_n^0$ of the algebra $\B_n^0$ generated by the
Jucys--Murphy elements $d_j$, $2\le j\le n$.
\end{problem}
\begin{ex}\label{e10.1} {\rm Let us consider the algebra $\B_3^0$.
This is an algebra over $\Q$ with generators $X_{12},X_{13},X_{23}$
subject the following quadratic relations
\begin{eqnarray}
&& X_{12}^2=X_{13}^2=X_{23}^2=0, \label{10.2}\\
&& X_{13}X_{12}=X_{12}X_{13}-X_{13}X_{23}+X_{23}X_{13}, \label{10.3}\\
&& X_{23}X_{12}=X_{12}X_{23}-X_{23}X_{13}+X_{13}X_{23}.~~~~~~~~~~~~~
~~~~~~~~~~~~\label{10.4}
\end{eqnarray}
Using these relations, we can find the new relations in the algebra
$\B_3^0$, namely,
\begin{eqnarray}
&& X_{23}X_{13}X_{23}-X_{13}X_{23}X_{13}+X_{12}X_{13}X_{23}-
X_{12}X_{23}X_{13}=0, \label{10.5}\\
&& X_{23}X_{13}X_{23}X_{13}+X_{12}X_{13}X_{23}X_{13}=0, \label{10.6}\\
&& X_{13}X_{23}X_{13}X_{23}+X_{12}X_{23}X_{13}X_{23}=0, \label{10.7}\\
&& X_{12}X_{23}X_{13}X_{23}-X_{12}X_{13}X_{23}X_{13}=0. \label{10.8}
\end{eqnarray}
\begin{rem}\label{r10.1} {\rm To be more precise, from the relations
(\ref{10.2})--(\ref{10.4}) follow only that
$$2\cdot{\rm LHS}(9.5)=0
$$
We expect that the $\Z$--form of the quadratic algebra $\B_n^0$ has only
2--torsion.}
\end{rem}
It is easy to check from the relations (\ref{10.2})--(\ref{10.8}) that
the Hilbert polynomial of the algebra $\B_3^0$ is equal to
$$H(\B_3^0;t)=1+3t+4t^2+3t^3+t^4=(1+t)^2(1+t+t^2)=H(\E_3^0;t).
$$
Now, let us consider the commutative subalgebra $\K_3^0:=\Q [d_2,d_3]$,
where $d_2=X_{12}$ and $d_3=X_{13}+X_{23}$. The Jucys--Murphy elements
$d_2$ and $d_3$ satisfy the following relations
$$d_2^2=0,\ \ d_2d_3=d_3d_2, \ \ (d_2+d_3)d_3^3=0,
$$
and the elements $1,d_2,d_3,d_2d_3,d_3^2,d_2d_3^2,d_3^3,d_2d_3^3$ form a
linear basis in the algebra $\K_3^0$. Thus we have
$$\dim\K_3^0=8, \ \ H(\K_3^0;t)=(1+t)^2(1+t^2).
$$
For general $n$ we expect that
\begin{equation}
H(\K_n^0;t)=\prod_{k=1}^{n-1}\frac{1-t^{2k}}{1-t}=H(W(B_{n-1});t),
\end{equation}
where $W(B_{n-1})$ is the Weyl group of type $B_{n-1}$.}
\end{ex}
\begin{con}\label{c10.2} The quadratic algebras $\B_n^0$ and $\E_n^0$ have
the same Hilbert polynomials
\begin{equation}
H(\B_n^0;t)=H(\E_n^0;t). \label{10.9}
\end{equation}
\end{con}
We can check (\ref{10.9}) for $n=3$ (Example~\ref{e10.1}) and $n=4$. In
order to prove the equality (\ref{10.9}) for $n=4$, we use the same method
as in Example~\ref{e6.1}. But, instead of identity (\ref{6.6}) in the
algebra $\E_n^0$, we have to use the following 14--terms relations in the
algebra $\B_n^0$, $n\ge 4$:
\begin{eqnarray}
&& X_{23}(X_{14}X_{24}X_{34}-X_{24}X_{14}X_{34}-X_{34}X_{14}X_{24}+
X_{34}X_{24}X_{14}) \label{10.10}\\
&&+(X_{13}X_{23}-X_{23}X_{13})(X_{24}X_{34}-X_{34}X_{24})
-X_{14}X_{24}X_{34}X_{24} \nonumber\\
&&+X_{24}X_{34}X_{24}X_{14}+X_{24}X_{14}X_{34}X_{24}
-X_{24}X_{34}X_{14}X_{24}\nonumber\\
&&+X_{34}X_{14}X_{24}X_{34}-X_{34}X_{24}X_{14}X_{34}=0, \nonumber \\
\nonumber\\
&&X_{12}(X_{14}X_{34}X_{24}-X_{24}X_{14}X_{34}-X_{34}X_{14}X_{24}+
X_{24}X_{34}X_{14}) \label{10.11}\\
&&+(X_{13}X_{23}-X_{23}X_{13})(X_{14}X_{24}-X_{24}X_{14})
+X_{14}X_{24}X_{34}X_{14} \nonumber\\
&&-X_{14}X_{34}X_{24}X_{14}-X_{14}X_{24}X_{14}X_{34}
+X_{24}X_{14}X_{34}X_{24}\nonumber\\
&&-X_{24}X_{34}X_{14}X_{24}+X_{34}X_{14}X_{24}X_{14}=0. \nonumber
\end{eqnarray}
It is also natural to ask whether or not the quadratic algebra $\B_n^0$
has the Koszul property (see, e.g. [Ma] and references therein).
%\newpage
\section{Quadratic algebra $\A_n^t$ }
\label{sec: quadr alg}
\neweq
In this section we consider the quadratic algebra $\A_n^t$ (denoted by
$\E C_n$ in [FK], Section~4.3) which is the
commutative quotient of the algebra $\E_n^t$.
\begin{de}\label{d7.1} Define the algebra $\A_n^t$ (of type $A_{n-1}$) as
the quadratic algebra over $\Z$ with generators $[ij]$, $1\le i<j\le n$,
which satisfy the following relations
\begin{eqnarray}
&&[ij][kl]=[kl][ij], \ \ {\rm for~ all} \ i,j,k,l; \label{7.1}\\
&&[ij][jk]=[ik][ij]+[ik][jk], \ \ i<j<k; \label{7.2}\\
&&[ij]^2=t_{ij}. \label{7.3}
\end{eqnarray}
\end{de}
Let us denote by $\A_n^0$
the specialization $t_{ij}=0$, $1\le i<j\le n$, of the algebra $\A_n^t$.
\begin{theorem}\label{t7.1} {\rm (cf. [A])} The Hilbert polynomial of the
algebra $\A_n^0$ is given by
\begin{equation}
H(\A_n^0;t)=(1+t)(1+2t)\cdots (1+(n-1)t). \label{7.4}
\end{equation}
\end{theorem}
\begin{cor}\label{c7.1} {\rm (cf. [A])} An additive basis of the
algebra $\A_n^t$ is given by the following set of (commutative) monomials
$$[j_1k_1]\ldots [j_lk_l],
$$
where $j_s<k_s$, $1\le s\le l$, and $k_1<k_2<\cdots <k_l$.
\end{cor}
The formula (\ref{7.4}) was stated at first in [FK], Proposition~4.2.
\begin{problem} It is an interesting task to find the Koszul dual of the
algebra $\A_n^t$.
\end{problem}
Let us postpone a proof of Theorem~\ref{t7.1} to the end of this Section,
and start with a review of some results obtained by V.I.~Arnold [A],
I.M.~Gelfand and A.N.~Var\-chen\-ko [GV].
It seems to be an interesting problem to understand the connections
between the algebra
$\E_n^0$ and the Orlik--Solomon algebra [OS] that corresponds to the
Coxeter hyperplane arrangement (of type $A_{n-1}$). Let us remind that
V.I.~Arnold [A] described this algebra as the quotient of the exterior
algebra of the vector space spanned by the $[ij]$ by the ideal generated
by the relations (\ref{7.2}). The "even analog" of the Orlik--Solomon
algebra of a hyperplane arrangement was introduced and studied by
I.M.~Gelfand and A.N.~Varchenko in [GV]. For
the Coxeter hyperplane arrangement of type $A_{n-1}$, the results of
Theorems~5 and 6 of [GV] can be formulated as follows.
Let ${\cal P}_n$ be
the ring of integer--valued functions that are defined on the complement
$M_C$ of the union of a finite set $C$ of hyperplaines in real,
$n$--dimensional affine space, that have constant values on each
connected component. The ring ${\cal P}_n$ has a distinguish set of
generators, namely, the Heaviside functions of hyperplaines: for each
hyperplane $f(x)=0$, the Heaviside function of the hyperplane $f$ takes
the value 1 from one side (where $f(x)>0$), and 0 from the other (where
$f(x)<0$). Any element of the ring ${\cal P}_n$ is a polynomial in the
Heaviside functions. Now let us assume that
$$\M_C=\R^n\setminus\bigcup_{1\le i<j\le n}\{ z_i-z_j=0\} .
$$
Let $X_{ij}$ be the Heaviside function of the hyperplane $H_{ij}=\{
z_i-z_j=0\}\subset\R^n$, then ([GV], Theorems~5, 6, and 8)
$\bullet$ The Heaviside functions $X_{ij}$, $1\le i<j\le n$, satisfy the
following relations:
\begin{eqnarray}
& & X_{ij}^2-X_{ij}=0, \label{7.5}\\
& & X_{ij}(X_{ik}-1)X_{jk}-(X_{jk}-1)X_{ik}(X_{ij}-1)=0, \label{7.6}
\end{eqnarray}
for any $1\le i<j<k\le n$.
$\bullet$ Let ${\cal T}_n$ be an ideal in the ring of polynomials $\Z
[X_{ij}, \ 1\le i<j\le n]$ generated by the left--hand sides of the
relations (\ref{7.5}) and (\ref{7.6}). Then the natural homomorphism $\Z
[X_{ij}]/{\cal T}_n\to{\cal P}_n$ is an isomorphism.
$\bullet$ Let us define the degree of a function $p\in{\cal P}_n$ as the
minimum of the degrees of polynomials in $\{ X_{ij}\}$ that represent
$p$, and denote by ${\cal P}_n^k$ the subspace of functions degree not
higher then $k$. Then there exists a noncanonical linear map
$$\pi_k~:~{\cal P}_n^k\to H^k(M_C,\Z ),
$$
which induces an isomorphism (${\cal P}_n^{-1}=\phi$)
$${\cal P}_n^k/{\cal P}_n^{k-1}~{\buildrel\sim\over\rightarrow}~H^k(M_C,\Z ),
\ 0\le k\le n.
$$
As a corollary, we have
\begin{equation}
H({\cal P}_n;t):=\sum_{k=0}^nt^k\dim{\cal P}_n^k/{\cal P}_n^{k-1}=
\prod_{j=1}^{n-1}(1+jt). \label{7.7}
\end{equation}
Now we are ready to prove Theorem~\ref{t7.1}.
{\it Proof of Theorem~\ref{t7.1}.} Let us rewrite the relations
(\ref{7.5}) and (\ref{7.6}) in slightly different form
\begin{eqnarray}
&& X_{ij}^2=\beta X_{ij}, \ \ 1\le i<j\le n; \label{7.8}\\
&& X_{ij}X_{jk}=X_{jk}X_{ik}+X_{ik}X_{ij}-\beta X_{ik},
\ \ 1\le i<j<k\le n, \label{7.9}
\end{eqnarray}
where $\beta$ is an additional deformation parameter.
We define a deformation ${\cal P}_{n,\beta}$ of the algebra ${\cal P}_n$
as the quotient algebra
$$\Z [X_{ij}]/{\cal T}_{n,\beta}={\cal P}_{n,\beta},
$$
where ${\cal T}_{n,\beta}$ is an ideal in the commutative ring of
polynomials $\Z [X_{ij}]$ generated by the relations (\ref{7.8}) and
(\ref{7.9}).
It is clear that ${\cal P}_{n,\beta =0}\cong\A_n^0$, and ${\cal
P}_{n,\beta}$ is a flat deformation of $\A_n^0$. Thus, we have
%(proof of Theorem~\ref{t7.1}):
$$\prod_{j=1}^{n-1}(1+jt)=H({\cal P}_n;t)=H({\cal P}_{n,\beta};t)=
H({\cal P}_{n,\beta =0};t)=H(\A_n^0;t).
$$
\qed
%\section{Quadratic algebra $\L_{n,\beta}$ }
%\label{sec: last}
%\neweq
\vskip 0.4cm
Let us introduce a noncommutative version of the algebra
${\cal P}_{n,\beta}$.
\begin{de}\label{d7.2} Define the algebra ${\cal L}_{n,\beta}$ (of type
$A_{n-1}$) as the quadratic algebra over $\Z [\beta ]$ with generators $[ij]$,
$1\le i<j\le n$, which satisfy the following relations
\begin{equation}
\label{7.10}
\begin{array}{rl}
&[ij][jk]=[jk][ik]+[ik][ij]+\beta [ik]\,,\\[.1in]
&[jk][ij]=[ik][jk]+[ij][ik]+\beta [ik],
\textrm{\ \ for } i<j<k\,;\quad\ \ \\[.1in]
\end{array}
\end{equation}
\begin{equation}
\label{7.11}
\begin{array}{rl}
&~[ij][kl]=[kl][ij],\\[.1in]
&~\textrm{whenever } \{i,j\}\cap\{k,l\}=\phi\,,
i<j, \textrm{ and }k<l \,. ~~~~~~~~~~
\end{array}
\end{equation}
\end{de}
The algebra ${\cal L}_{n,\beta}$ is a smooth deformation of the algebra
$\G_n$, and has the same Hilbert series:
$$H({\cal L}_{n,\beta};t):=\sum_{k=0}^{\infty}t^k\dim{\cal
L}_{n,\beta}^{(k)}/{\cal L}_{n,\beta}^{(k-1)}=H(\G_n;t).
$$
We can define also the quotients $\L_{n,\beta}^t$, $\L_{n,\beta}^0$, and
commutative subalgebras $\H_{n,\beta}$, $\H_{n,\beta}^t$ and
$\H_{n,\beta}^0$ generated by the Dunkl elements $\theta_j$ (see
(\ref{2.1})), $1\le j\le n$.
It seems interesting to describe the
relations in the commutative subalgebra
$\H_{n,\beta}^t\subset\L_{n,\beta}^t$. The work is in progress and we
hope to present our results in the nearest future.
\newpage
\begin{thebibliography}{AKL}
\setlength{\itemsep}{0.05in}%{-.02in}
\bibitem [A] {} Arnold V.I., {\it The cohomology ring of the pure braid
group,} Mat. Zametki, 1969, v.5, 227-231;
\bibitem [B] {} Brieskorn E., {\it Automorphic sets, and braids and
singularities,} Contemporary Math., 1988, v.78, p.45-115;
\bibitem [B-N] {} Bar--Natan D., {\it Bibliography of Vassiliev invariants;}
This bibliography is available electronically at
\texttt{http://www.ma.huji.ac.il/$\sim$drorbn};
\bibitem [Ch1] {} Cherednik I., {\it Double affine Hecke algebras and
Macdonald conjectures,} Annals of Math., 1995, v.141, p.191-216;
\bibitem [Ch2] {} Cherednik I., {\it Lectures on affine quantum
Knizhnik--Zamolodchikov equations, quantum many body problems, Hecke
algebras, and Macdonald theory,} Lectures given at the IIAS, Kyoto, 1996;
\bibitem [Co] {} Cohen F.R., {\it Artin's braid group, classical homotopy
theory, and sundry other curiosities,} Contemporary Math., 1988, v.78,
p.167-206;
\bibitem [Dr1] {} Drinfeld V.G., {\it On quasi--Hopf algebras,} Leningrad
Math. Journ., 1990, v.1, p.1419-1457;
\bibitem [Dr2] {} Drinfeld V.G., {\it On quasi--triangular quasi--Hopf
algebras and a group closely connected to $Gal(\overline\Q/\Q)$,}
Leningrad Math. Journ., 1991, v.2, p.829-860;
\bibitem [Du] {} Dunkl C.F., {\it Differential--difference operators
associated to reflection groups,} Trans. Amer. Math. Soc., 1989, v.311,
p.167-183;
\bibitem [FK] {} Fomin S. and Kirillov A.N., {\it Quadratic algebras,
Dunkl elements and Schubert calculus,} Prepr., 1997,
AMSPPS \#199703--05--001, 34p.;
\bibitem [F] {} Funar L., {\it Vassiliev invariants I: braid groups and
rational homotopy theory,} Prepr., 1995, q-alg/9510008, 24p.;
\bibitem [GV] {} Gelfand I.M. and Varchenko A.N., {\it Heaviside functions
of a configuration,} Funkts. Anal. Prilozhen., 1987, v.21, n.4. p.1-18;
\bibitem [H] {} Heckman G.J.,
{\it A remark on the Dunkl differential-difference operators,
\emph{in}: Harmonic analysis on reductive groups,}
W.~Barker and P.~Sally, eds.,
Birkh\"auser, 1991, 181--191;
\bibitem [J] {} Jucys A.A., {\it On the Young operators of symmetric
groups,} Lithuanian Physics Journal, 1966, v.6, p.163-180;
\bibitem [Ki] {} Kirillov A.A., Jr.,
{\it Lectures on affine Hecke algebras and Macdonald's conjectures,}
Prepr., 1995, 43p.;
\bibitem [K] {} Kirillov A.N., {\it Cauchy identities for universal
Schubert polynomials,} Prepr., 1997, q-alg/9703047, 16p.;
\bibitem [KN] {} Kirillov A.N. and Noumi M., {\it Affine Hecke algebras
and raising operators for Macdonald polynomials,}
to appear in Duke Math. Journ.; q-alg/9605004, 1996, 35p.;
\bibitem [Ko1] {} Kohno T., {\it S\'eries de Poincar\'e--Koszul
associ\'ee aux groupes de tresses pures,} Invent. Math., 1985, v.82.
p.57-75;
\bibitem [Ko2] {} Kohno T., {\it Linear representations of braid groups
and classical Yang--Baxter equations,} Contemporary Math., 1988, v.78,
p.339-363;
\bibitem [Ko3] {} Kohno T., {\it Vassiliev invariants and de Rham complex
on the space of knots,} Contemporary Math.,1994, v.179, p.123-138;
\bibitem [L] {} Lin X.-S., {\it Braid algebras, trace modules and
Vassiliev invariants,} Prepr., 1993, 28p.;
\bibitem [Ma] {} Manin Yu. I.,
{\it Quantum groups and noncommutative geometry},
Universit\'e de Montr\'eal,
Centre de Recherches Math\'ematiques, Montr\'eal, 1988.
\bibitem [Mu] {} Murphy G.E., {\it A new construction of Young's
seminormal representation of the symmetric groups,} J. of Algebra, 1981,
v.69, p.287-297;
\bibitem [OS] {} Orlik P. and Solomon L., {\it Combinatorics and topology
of complements of hyperplanes,} Invent. Math., 1980, v.56, p.167-189;
\bibitem [P] {} Postnikov A.,
{\it On Quantum Monk's and Pieri's Formulas,}
Prepr., 1997, AMSPPS \#199703-05-002, 12p;
\bibitem [Ra] {} Ram A., {\it Seminormal representations of Weyl groups
and Iwahori--Hecke algebras,} 1995, to appear in Proc. London Math. Soc.;
\bibitem [Ro] {} Roos J.-E., {\it Some non-Koszul algebras,} Prepr.,
1997, 5p.
\end{thebibliography}
\end{document}