Fractional statistics in some exactly solvable Calogero-like models with PT invariant interactions
Extracted source text ยท arXiv cond-mat/0201074
\documentstyle[12pt]{article}
\setlength {\topmargin}{-12mm}
\textheight=227mm
\textwidth=157mm
%
\def\beq{\begin{equation}}
\def\eeq{\end{equation}}
\def\bea{\begin{eqnarray}}
\def\eea{\end{eqnarray}}
\def\nn{\nonumber}
\def\ba{\begin{array}}
\def\ea{\end{array}}
\def\d{\partial}
\def\v{\vert}
\def\l{\langle}
\def\r{\rangle}
%
\def\one{1\hskip -1mm{\rm l}}
%
\renewcommand{\thefootnote}{\fnsymbol{footnote}}
\setcounter{footnote}{1}
%
\begin{document}
\baselineskip17pt
\begin{center}
{\large \bf \sf
Fractional statistics in some exactly solvable \\ Calogero-like
models with PT invariant interactions }
\vspace{1.5cm}
{\sf B. Basu-Mallick},
\footnote{E-mail address~: biru@theory.saha.ernet.in}
\bigskip
{\em Saha Institute of Nuclear Physics, Theory Group \\
1/AF Bidhan Nagar, Kolkata 700 064, India }
\bigskip
\end{center}
\vspace {1.75 cm}
\baselineskip=15.5pt
\noindent {\bf Abstract }
Here we review a method for constructing exact eigenvalues and eigenfunctions
of a many-particle quantum system, which is obtained by adding some
nonhermitian but PT invariant (i.e., combined parity and time reversal
invariant) interaction to the Calogero model. It is shown that such extended
Calogero model leads to a real spectrum obeying generalised exclusion
statistics. It is also found that the corresponding exchange statistics
parameter differs from the exclusion statistics parameter and exhibits a
`reflection symmetry' provided the strength of the PT invariant interaction
exceeds a critical value.
\vspace {.05 cm}
\newpage
\noindent \section {Introduction }
\renewcommand{\theequation}{1.{\arabic{equation}}}
\setcounter{equation}{0}
\medskip
It is well known that integrable dynamical models and spin chains with
long range interactions exhibit fractional statistics or generalised exclusion
statistics (GES) \cite{ha},
which is believed to play an important role in many strongly correlated
systems of condensed matter physics.
The $A_{N-1}$ Calogero model (related to $A_{N-1}$ Lie algebra)
is the simplest example of such dynamical model, containing $N$
particles on a line and with Hamiltonian given by
\cite {ca,su}
\beq
H= - {1\over 2} \sum_{j=1}^N {\d^2 \over \d x_j^2}
+ {\omega^2\over 2} \sum_{j=1}^N x_j^2 +
{g \over 2} \sum_{j\neq k} {1 \over (x_j -x_k)^2} \, ,
\label{1}
\eeq
where $g$ is the coupling constant associated with long-range interaction.
One can exactly solve this Calogero model and find out the complete set of
energy eigenvalues as
\beq
E_{n_1,n_2, \cdots , n_N} =
{N \omega \over 2} \left [ 1 + (N-1)\nu \right ] + \omega \sum_{j=1}^N n_j .
\label {2}
\eeq
Here $n_j$s are non-negative integer valued
quantum numbers with $n_j \leq n_{j+1}$ and $\nu $
is a real positive parameter which is related to $g$ as
\beq
g = \nu^2 - \nu \, .
\label {3}
\eeq
It may be noted that, apart from a constant shift for all energy levels,
the spectrum (\ref {2}) coincides with that of $N$ number of free
bosonic oscillators. Furthermore, one can easily remove the above mentioned
constant shift for all energy levels and express (\ref{2})
exactly in the form of energy eigenvalues for free oscillators:
$E_{n_1,n_2, \cdots , n_N} = {N \omega \over 2} + \omega
\sum_{j=1}^N \bar n_j,
$
where ${\bar n_j} = n_j + \nu ( j-1) $ are quasi-excitation numbers.
However it is evident that these ${\bar n_j} $s
are no longer integers and they
satisfy a modified selection rule given by
${\bar n_{j+1}} - {\bar n_j} \geq \nu $,
which restricts the difference
between the quasi-excitation numbers to be at least
$ \nu$ apart.
As a consequence, the
Calogero model (\ref{1}) provides a microscopic
realisation for fractional statistics with $\nu $ representing
the corresponding GES parameter \cite{is,mu,po,bk}.
Recently, theoretical investigations on
different nonhermitian Hamiltonians have received a major
boost because many such systems,
whenever they are invariant under combined parity
and time reversal (PT) symmetry, lead to real energy
eigenvalues \cite{pt1,pt2,pt3,pt4}.
This seems to suggest that the condition of hermiticity
on a Hamiltonian can be replaced by the weaker condition of PT symmetry
to ensure that the corresponding eigenvalues would be real ones.
However, till now this is merely a conjecture supported by several examples.
Moreover, in almost all of these examples, the Hamiltonians of
only one particle in one space dimension have been considered.
Therefore, it should be interesting to test
this conjecture for the cases of nonhermitian $N$-particle Hamiltonians
in one dimension which remain invariant
under the $PT$ transformation \cite{bm}:
\beq
i \rightarrow -i, \ \ x_j \rightarrow - x_j, \ \ p_j \rightarrow p_j \, ,
\label {4}
\eeq
where $j \in [1,2, \cdots , N],$ and
$x_j$ ($p_j \equiv -i \frac {\partial }{\partial x_j}$) denotes
the coordinate (momentum) operator of the $j$-th particle.
In particular, one may construct an extension of Calogero model by
adding to it some nonhermitian but PT invariant interaction,
and enquire whether such extended model would lead to
real spectrum.
The purpose of the present article is to review the progress \cite{bk,bm}
on the above mentioned problem for some special cases,
where the PT invariant extension of Calogero model can
be solved exactly. In Sec.2 of this article we consider such a PT invariant
extension of $A_{N-1}$ Calogero model and show that,
within a certain range of the related parameters, this
extended Calogero model yields real eigenvalues. Next,
in Sec.3, we explore the connection of these real eigenvalues with
fractional statistics. Section 4 is the concluding section.
\noindent \section { Exact solution of an extended Calogero model}
\renewcommand{\theequation}{2.{\arabic{equation}}}
\setcounter{equation}{0}
Let us consider a Hamiltonian of the form \cite{bk}
\beq
{\cal H} = H + \delta \sum_{j\neq k} {1 \over x_j - x_k}{\d \over \d x_j},
\label {5}
\eeq
where $H $ is given by eqn.(\ref {1}) and
$\delta $ is a real parameter.
It may be observed that though the Hamiltonian
(\ref{5}) violates hermiticity property due to the presence of momentum
dependent term like
$ \delta \sum_{j\neq k} {1 \over x_j - x_k}{\d \over \d x_j} $, it
remains invariant under the combined PT transformation (\ref {4}).
Next we recall that,
$A_{N-1}$ and $B_N$ Calogero models as well as their distinguishable variants
have been solved recently by mapping them to a system of
free oscillators \cite{gp,so,bf,wa}.
With the aim of solving the PT invariant extension
(\ref{5}) of $A_{N-1}$ Calogero model through a similar produce,
we assume that (justification for this assumption will be given later)
the ground state wave function for this extended
model is given by
\beq
\psi_{gr}= e^{- {\omega\over 2} \sum_{j=1}^N x_j^2 }
\prod_{j<k} (x_j-x_k)^\nu ,
\label {6}
\eeq
where $\nu $ is a real positive number which is
related to the coupling constants
$g$ and $\delta $ as
\beq
g = \nu^2 - \nu ( 1+ 2 \delta ) \, .
\label {7}
\eeq
Now if we use the expression
(\ref {6}) for a similarity transformation to
the Hamiltonian (\ref {5}), it reduces to an `effective Hamiltonian'
of the form
\beq
{\cal H}'=\psi_{gr}^{-1} {\cal H} \psi_{gr} = S^- + \omega S^3 + E_{gr} \, ,
\label {8}
\eeq
where the Lassalle operator ($S^-$) and Euler operator ($S^3$) are given by
\beq
S^- = - {1\over 2} \sum_{j=1}^N {\d^2 \over \d x_j^2} -
(\nu- \delta)\sum_{j \neq k}
{1 \over x_j - x_k}{\d \over \d x_j}, ~~~
S^3 = \sum_{j=1}^N x_j { \d \over \d x_j } ,
\label {9}
\eeq
and
\beq
E_{gr}= {N \omega \over 2} \left [ 1 + (N-1)(\nu -\delta ) \right ] .
\label {10}
\eeq
It is easy to see that
the Lassalle operator and Euler operator, as defined
in eqn.(\ref {9}), satisfy the simple
commutation relation:
$ [S^3,S^-]= -2 S^-$. Using therefore the well known Baker-Hausdorff
transformation we can remove the $S^-$ part of the effective Hamiltonian
${\cal H}' $ and through some
additional similarity transformations reduce it finally to the
free oscillator model \cite{bk}
\beq
H_{free} = {\cal S}^{-1} \left({\cal H}'- E_{gr}\right) {\cal S}
= - {1\over 2} \sum_{j=1}^N {\d^2 \over \d x_j^2} +
{\omega^2 \over 2} \sum_{j=1}^N x_j^2 - {\omega N \over 2},
\label {11}
\eeq
where ${\cal S} =e^{ {1\over 2 \omega } S^- }
e^{ {1\over 4 \omega } \nabla^2 }
e^{ {\omega \over 2} \sum_{j=1}^N x_j^2 } $ and
$ \nabla^2 = \sum_{j=1}^N {\d^2 \over \d x_j^2} $.
Due to similarity transformations in (\ref {8}) and (\ref {11}),
one may naively think that
the eigenfunctions of the extended Calogero model (\ref {5}) can be
obtained from those of free oscillators as:
$ \psi_{n_1, n_2 ,\cdots, n_N} = \psi_{gr} {\cal S} \left\{ \prod_{j=1}^N
e^{ - {\omega \over 2} x_j^2 } H_{n_j}(x_j) \right \}$, where
$n_j$s are arbitrary non-negative integers and
$H_{n_j}(x_j)$ denotes the Hermite polynomial of order $n_j$.
However it is easy to check that,
similar to the case of $A_{N-1}$ Calogero model \cite{gp},
the action of ${\cal S}$ on free oscillator eigenfunctions
leads to a singularity unless they are symmetrised with respect to
all coordinates. Therefore, nonsingular eigenfunctions of
the extended Calogero model (\ref {5}) can be
obtained from the
eigenfunctions of free oscillators as
\beq
\psi_{n_1, n_2, \cdots , n_N} = \psi_{gr} \ {\cal S} \Lambda_+
\left\{ \prod_{j=1}^N
e^{ - {\omega \over 2} x_j^2 } H_{n_j}(x_j) \right \} \, ,
\label {12}
\eeq
where $\Lambda_+ $ completely symmetrises all coordinates and thus
projects the distinguishable many-particle wave functions to the
bosonic part of the Hilbert space. Evidently, the eigenfunctions
(\ref {12}) will be mutually independent if the excitation
numbers $n_j$s obey the bosonic selection rule: $n_{j+1} \geq n_j $.
Thus, in spite of the fact that the interacting Hamiltonian (\ref {5})
is convertible to the free oscillator model, the need for symmetrization
shows that the many-particle correlation is in fact inherent in this model.
The eigenvalues of the Hamiltonian (\ref{5})
corresponding to the states (\ref{12}) will
naturally be given by \cite{bk}
\beq
E_{n_1,n_2, \cdots , n_N} = E_{gr}+\omega \sum_{j=1}^N n_j=
{N \omega \over 2} \left [ 1 + (N-1)(\nu -\delta ) \right ] +
\omega \sum_{j=1}^N n_j .
\label {13}
\eeq
Since $\delta $ and $\nu $ are
real parameters, the energy eigenvalues
(\ref {13}) are also real ones. Thus we interesting find that
the nonhermitian PT invariant Hamiltonian (\ref {5}) yields a real
spectrum. Furthermore, it is evident that for all $n_j=0,$
the energy $E_{n_1,n_2, \cdots , n_N}$
attains its minimum value
$E_{gr}$. At the same time, as can be easily worked out
from eqn.(\ref {12}),
the corresponding eigenfunction reduces to
$\psi_{gr}$ (\ref {6}).
This proves that $\psi_{gr}$ is indeed the ground state wave function
for Hamiltonian (\ref{5}) with eigenvalue $E_{gr}$.
It may be observed that the eigenfunctions (\ref {12}) pick
up a phase factor $(-1)^\nu $ under the exchange of any two
particles. Therefore,
$\nu $ represents the exchange statistics parameter
for the extended Calogero model (\ref {5}). By solving the quadratic
eqn.(\ref {7}), one can explicitly write down $\nu $
as a function of $g$ and $\delta $ as
\beq \nu = (\delta + {1\over 2}) \pm
\sqrt{ g + (\delta +{1\over 2})^2 } .
\label {14}
\eeq
For the purpose of obtaining real
eigenvalues (\ref {13}) as well as nonsingular eigenfunctions
(\ref {12}) at the limit $x_i
\rightarrow x_j $,
we have assumed at the beginning of this section
that $\nu $ is a real positive parameter. This assumption
leads to a restriction on the allowed values of the coupling
constants $g$ and $\delta $ in the following way.
First of all,
for the case $g < - (\delta +{1\over 2})^2 $,
eqn.(\ref {14}) yields two imaginary solutions.
Secondly, for the case $\delta < - {1\over 2} , ~~
0 > g > - (\delta +{1\over 2})^2 $
eqn.(\ref {14}) yields two real but negative solutions.
Inequalities corresponding to these two cases represent two forbidden
regions of $( \delta ,g)$ plane which are excluded
from our analysis.
For the case $g>0$ with arbitrary value of $\delta $, one gets a real positive
and a real negative solution from eqn.(\ref {14}). The real positive
solution evidently leads to
physically acceptable set of eigenfunctions and corresponding
eigenvalues within this allowed region of $( \delta ,g)$ plane.
Finally we consider the parameter range
$\delta > - {1\over 2} , ~~ 0 > g > - (\delta +{1\over 2})^2 $,
for which eqn.(\ref {14}) yields two real positive solutions.
It is easy to see that these
two real positive solutions are related to each other through a
`reflection symmetry' given by $\nu \rightarrow 1 + 2\delta - \nu $.
Consequently, for each point on the
$( \delta , g )$ plane within this allowed parameter range,
one obtains two different values of
the exchange statistics parameter
leading to two distinct sets of physically acceptable
eigenfunctions and eigenvalues.
Thus we curiously find that a kind of `phase transition' occurs
at the line $\delta =
- {1 \over 2}$ on the $(\delta ,g)$ plane.
For the case $\delta > - {1 \over 2}$,
exchange statistics parameter
shows the reflection symmetry
when $g$ is chosen within an interval
$ - ( {1\over 2} + \delta )^2 < g \leq 0$.
On the other hand
for the case $\delta \leq - {1 \over 2}$,
such reflection symmetry is lost for any possible value of $g$.
We have seen in this section that,
similar to the case of $A_{N-1}$ Calogero model,
the extended model (\ref {5}) can also be solved by mapping it
to a system of free harmonic oscillators.
So it is natural to enquire whether
this extended model is directly related to the
$A_{N-1}$ Calogero model through some similarity transformation.
Investigating along this line \cite{bm}, we find that
\beq
\Gamma ^{-1}{\cal H} \Gamma = H^\prime =
\frac{ 1}{2}\sum p_j^2 + \frac{ 1}{2}\omega ^2
\sum x_j^2 + g^\prime
\sum_{j\neq k}^N \frac{1}{(x_j-x_k)^2} \, ,
\label{15}
\eeq
where $ \Gamma = \prod_{j< k} (x_j-x_k)^\delta $,
and $H^\prime $ denotes the Hamiltonian of $A_{N-1}$
Calogero model with `renormalised' coupling constant
given by $g^\prime = g + \delta (1+\delta ) $. Due to the existence of
such similarity transformation, one may expect that the Hamiltonians
${\cal H}$ and $H^\prime$ always lead to exactly same eigenvalues.
However it should be noted that,
within a parameter range given by
$\delta > 0 , ~g > - \delta (1+ \delta )$, there exists
a positive solution of eqn.(\ref {7}) satisfying the condition
$ \nu - \delta < 0 $. Therefore, we can not get any lower bound
for the corresponding energy eigenvalues (\ref {13}) at $N
\rightarrow \infty $ limit. On the other hand, the
energy eigenvalues (\ref {2}) of $A_{N-1}$ Calogero model are certainly
bounded from below for all possible choice of $N$ and $g$.
So there exists
a parameter range within which the spectrum of extended Calogero model differs
qualitatively from the spectrum of the original Calogero model.
To explain this rather unexpected result, we first observe that
the renormalised coupling constant
$ g^\prime $ would be a positive quantity within the
above mentioned parameter range.
Consequently, the corresponding exclusion statistics parameter
$\nu^\prime $, which is obtained by solving eqn.(\ref {3}),
has one positive and one negative solution.
One usually throws away this negative solution of
$ \nu^\prime $, since the corresponding eigenfunctions
become singular at the limit $x_j \rightarrow x_k$. However,
by using the relation (\ref {15}), such singular eigenfunctions
(denoted by $\psi^\prime(x_1,x_2, \cdots ,x_N$))
may now be used to construct the
eigenfunctions of extended Calogero model
(denoted by $\psi(x_1, x_2, \cdots x_N$)) as
\beq
\psi (x_1, x_2, \cdots , x_N ) ~=~
\prod_{j < k} (x_j-x_k)^\delta
\psi^\prime (x_1, x_2, \cdots , x_N ) \, .
\label{16}
\eeq
It can be easily checked that, due to the existence of the factor
$\prod_{j < k} (x_j-x_k)^\delta $, the r.h.s. of the above equation
becomes nonsingular at the limit $x_j \rightarrow x_k \ $.
Thus we curiously find that
singular eigenfunctions of $H^\prime $ can be used to generate nonsingular
eigenfunctions of ${\cal H}$ through the relation (\ref {16}).
This shows that the similarity transformation (\ref {15}) is a subtle one
and, within a certain parameter range, eigenvalues of extended
Calogero model will match with those of Calogero model (having
renormalised coupling constant) only if
the corresponding unphysical eigenfunctions are taken into account.
\vspace{.75cm}
\noindent \section {Connection with fractional statistics}
\renewcommand{\theequation}{3.{\arabic{equation}}}
\setcounter{equation}{0}
We have mentioned in Sec.1 that GES
can be realised microscopically in
$A_{N-1}$ Calogero model with hermitian Hamiltonian.
The GES parameter for this Calogero model
is a measure of `level repulsion' of the
quantum numbers generalising the Pauli exclusion principle.
Now for exploring GES in the case of PT invariant model (\ref {5}),
we observe that eqn.(\ref{13}) can be rewritten \cite{bk}
exactly in the form of energy spectrum for $N$ free oscillators as
\beq
E_{n_1,n_2\cdots n_N} = \frac{N \omega }{2} + \omega
\sum_{j=1}^N \bar{n}_j \, ,
\label{17}
\eeq
where
\beq
\bar{n}_j = n_j + (\nu - \delta ) (j-1)
\label {18}
\eeq
are quasi-excitation numbers.
However, from eqn.(\ref {18}) it is evident
that such quasi-excitation numbers are no longer integers
and satisfy a modified selection rule:
$ \bar{n}_{j+1}- \bar{n}_j \geq \nu - \delta \, .$ Since the minimum
difference between two consecutive
$\bar{n}_j $s is given by
\beq
\tilde{\nu } = \nu - \delta_ ,
\label {19}
\eeq
the spectrum of extended $A_{N-1}$ Calogero model (\ref {5})
satisfies GES with parameter $ \tilde{\nu}$ \cite{bk}.
Several comments about this GES
parameter are in order.
It may be to noted that for $\delta \neq 0$, the GES parameter
$\tilde \nu $ is different from the power
index $ \nu $, which is responsible for the
symmetry of the wave function.
Therefore we may interestingly conclude that unlike Calogero model,
the exclusion statistics for model
(\ref{5}) differs from its exchange statistics. Furthermore it is already
noticed that, on a region of
$(\delta ,g)$ plane satisfying the inequalities
$\delta > 0 , ~g > - \delta (1+ \delta )$, there exists
a positive solution of eqn.(\ref {7}) which yields a negative
value of $\tilde \nu $. For this case, however, one does not
get well defined thermodynamic
relations at $N\rightarrow \infty$ limit and, therefore,
can not interpret $\tilde \nu $ as the GES parameter.
By using eqn.(\ref {7})
and (\ref {19}), we find the relation
\beq
\tilde{\nu }^2 - \tilde{\nu } = g + \delta ( \delta + 1) ,
\label {20}
\eeq
which clearly describes a parabolic
curve in the coupling constant plane $(\delta , g) $ for any fixed value
of $\tilde { \nu} $.
As a consequence of this, the competing effect of the independent
coupling constants $g$ and $\delta $ can make
the GES feature of (\ref{5}) much
richer in comparison with the Calogero model.
For example, while
bosonic (fermionic) excitations in the Calogero model
occur only in the absence of long-range interaction,
the quasi-excitations in (\ref{5}) can behave as
pure bosons (fermions)
even in the presence of both the long-range
interactions satisfying the constraint
$\tilde \nu(\delta ,g)=0$ \ ($\tilde \nu(\delta ,g)=1$).
Both of these constraints
lead to
the same parabolic curve $g=-\delta (1+\delta )$.
A family of such parabolas with shifted apex points are generated
for other values of
$\tilde \nu$ and the lowest apex point is attained at
$\tilde \nu= {1\over 2}$,
where
the quasi-excitations would behave as semions.
\noindent \section{Conclusion}
Here we construct a many-particle
quantum system (\ref {5}) by adding some nonhermitian
but combined parity and time reversal (PT)
invariant interaction to the $A_{N-1}$ Calogero model.
By using appropriate similarity transformations, we are able
to map this extended Calogero model to a set of free harmonic oscillators
and solve this model exactly. It turns out that this
many-particle system with nonhermitian Hamiltonian
yields a real spectrum. This fact supports the conjecture that
the condition of hermiticity
on a Hamiltonian can be replaced by the weaker condition of PT symmetry
to ensure that the corresponding eigenvalues would be real ones.
It is also found that the spectrum of extended
Calogero model obeys a selection rule which leads to generalised
exclusion statistics (GES).
However, this extended Calogero model exhibits some remarkable properties
which are absent in the case of usual
Calogero model. For example, we curiously find that the
GES parameter for this extended Calogero model
differs from the corresponding exchange statistics parameter. Moreover
a `reflection symmetry' of the
exchange statistics parameter, which is known to exist for
$A_{N-1}$ Calogero model, can be found
in the case of extended model only if
the strength of PT invariant interaction exceeds a critical value.
Finally we note that,
it is possible to obtain another exactly solvable many-particle
quantum system by adding some nonhermitian
but PT invariant interactions to the $B_N$ Calogero model (associated
with $B_N$ Lie algebra) \cite{bm}. Such a PT invariant
model also leads to real spectrum with
properties quite similar to the case of
extended $A_{N-1}$ Calogero model.
\bigskip
\noindent {\bf Acknowledgments }
I would like to thank Prof. A. Kundu and Dr. B.P. Mandal for
collaborating with me in Ref.7 and Ref.12. I also like to
thank Prof. M.L. Ge for kindly inviting me to the
`APCTP-Nankai Symposium', October 7-10, 2001, Tianjin, China.
\begin{thebibliography}{99}
\bibitem{ha}F. D. M. Haldane, {\em Phys. Rev. Lett.} {\bf 67} (1991) 937.
\bibitem{ca}F. Calagero, {\em J. Math. Phys.} {\bf 10} (1969) 2191;
{\em J. Math. Phys.} {\bf 12} (1971) 419.
\bibitem{su}B. Sutherland, {\em J. Math. Phys.} {\bf 12} (1971) 246.
\bibitem{is}S.B. Isakov, {\em Mod. Phys. Lett. } {\bf B 8} (1994) 319;
{\em Int. J. Mod. Phys. } {\bf A 9} (1994) 2563.
\bibitem{mu}M.V.N. Murthy and R. Shankar, {\em Phys. Rev. Lett.}
{\bf 73} (1994) 3331.
\bibitem{po} A.P. Polychronakos, in
{\em Les Houches Lectures (Summer 1998)}, hep-th/9902157.
\bibitem{bk}B. Basu-Mallick and A. Kundu,
{\em Phys. Rev. } {\bf B 62} (2000) 9927.
\bibitem{pt1}C.M. Bender and S. Boettcher, {\em Phys. Rev. Lett.}
{\bf 80} (1998) 5243; {\em J. Phys. } {\bf A 31} (1998) L273.
\bibitem{pt2}C.M. Bender, S. Boettcher and P.N. Meisinger,
{\em J. Math. Phys.} {\bf 40} (1999) 2210.
\bibitem{pt3} F.M. Fernandez, R. Guardiola, J. Ros and M. Znojil,
{\em J. Phys. }{\bf A 32} (1999) 3105.
\bibitem{pt4} A. Khare and B. P. Mandal,
{\em Phys. Lett. } {\bf A 272} (2000) 53.
\bibitem{bm}B. Basu-Mallick and B.P. Mandal,
{\em Phys. Lett. } {\bf A 284} (2001) 231.
\bibitem{gp}N. Gurappa and P.K. Panigrahi,
{\em Phys. Rev.} {\bf B 59} (1999) R2490.
\bibitem{so}K. Sogo, {\em J. Phys. Soc. Jpn.} {\bf 65} (1996) 3097.
\bibitem{bf}T.H. Baker and P.J. Forrester, {\em Nucl. Phys.}
{\bf B 492} (1997) 682.
\bibitem{wa}H. Ujino, A. Nishino and M. Wadati,
{\em J. Phys. Soc. Jpn.} {\bf 67} (1998) 2658.
\end{thebibliography}
\end{document}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%