Large-N expansion based on the Hubbard operator path integral representation and its application to the t-J model II. The case for finite $J$
Extracted source text ยท arXiv cond-mat/0409748
%\documentstyle[epsfig,a4,12pt]{article}
%\documentstyle[twocolumn,epsfig,a4,12pt]{article}
%\documentstyle[twocolumn,epsfig,prb,aps,amsmath,amssymb]{revtex}
%\documentstyle[twocolumn,epsfig,prb,aps,amsmath,]{revtex}
\documentclass[prb,twocolumn,amsmath,amssymb,showpacs]{revtex4}
\usepackage{graphicx}
\def\beq{\begin{equation}}
\def\eeq{\end{equation}}
\def\beqa{\begin{eqnarray}}
\def\eeqa{\end{eqnarray}}
\def\rmd{{\rm d}}
\begin{document}
%\twocolumn[\hsize\textwidth\columnwidth\hsize\csname
%@twocolumnfalse\endcsname
\title{
Large-$N$ expansion based on the Hubbard operator
path integral representation
and its application to the t-J model II. The case for finite $J$.
}
\author{
Adriana Foussats and Andr\'es Greco
}
\affiliation{
Facultad de Ciencias Exactas, Ingenier\'{\i}a y Agrimensura
and Instituto de F\'{\i}sica Rosario
(UNR-CONICET).
Av. Pellegrini 250-2000 Rosario-Argentina.
}
\date{\today}
\begin{abstract}
We have introduced a new perturbative approach for $t-J-V$ model
where Hubbard operators are treated as
fundamental objects.
Using our vertices and propagators we have developed
a controllable large-$N$ expansion
to calculate different correlation functions. We have investigated
charge density-density response and the phase diagram of
the model.
The charge correlations functions are not very sensitive to the value
of $J$ and they show collective peaks (or zero sound) which are more
pronounced when they are well separated (in energy) from the
particle-hole continuum. For a given $J$ a Fermi liquid state is
found to be stable for doping
$\delta$ larger than a critical doping $\delta_c$. $\delta_c$ decreases
with decreasing $J$. For the physical region of the parameters
and, for $\delta< \delta_c$, the system enters
in an incommensurate flux or
DDW phase.
The inclusion of the nearest-neighbors Coulomb repulsion $V$ leads to a
CDW phase when $V$ is larger than a critical value $V_c$.
The dependence of $V_c$
with $\delta$ and $J$ is shown.
We have compared the results with other ones in the literature.
\end{abstract}
\pacs{ 71.10.-w,71.27.+a}
\maketitle
%\vskip2pc]
%=====================================================================
% BODY OF PAPER
\section{Introduction}
In the last years a large part of the solid state community has been devoted
to understand the physics of strongly electronic correlated models.
The importance of this study is supported by
clear experimental facts.
High-$T_c$ superconductors \cite{Anderson} and organic superconductors
\cite{McKenzie} are considered, for instance, electronic correlated
systems.
They show that
electronic properties are very different from those expected in usual metals.
For better understanding of the physics of these systems,
it is very important
to develop different methods for studying models for correlated electrons.
The two dimensional ($2D$) $t-J$ model is probably one of the most simple
models we should understand in the first place.
The $t-J$ model is the strong coupling version of the Hubbard model
\cite{Izyumov}. In spite of its simple form, there is no exact solution
for this model
until now and,
many analytical and numerical \cite{Dagotto} methods
have been developed and the obtained results confronted to each other.
In the $t-J$ model double occupancy is forbidden and,
the model can be written
in terms of the correlated or projected operators, also known as
Hubbard X-operators \cite{Hubbard}.
Hubbard operators verify complicated commutation rules
which are very different from the familiar commutation rules for usual
fermions and bosons.
One of the common methods, used for avoiding this problem, introduces
slave-particles\cite{Izyumov} in order to decouple the original $X$-operator.
These slave-particles verify usual
commutation rules but, they are fictitious and
sometimes it is not clear if the obtained results are
genuine or artifacts from merely decoupling. On the other hand,
the decoupling scheme introduces
a gauge degree of freedom which requires a gauge fixing.
The gauge fixing is a long discussed fundamental problem
(see for example Ref.[\onlinecite{Arrigoni}]).
Another problem in the treatment of the $t-J$ model is
the absence of any small
coupling parameter suitable for a perturbative expansion.
To deal with this problem, strong-coupling techniques have been developed.
Between them,
what will be important for the present paper is the large-$N$
expansion where $N$ is the number of electronic degrees of freedom per
site (see below).
The large-$N$ expansion has been used extensively in the context of
slave-boson approach \cite{Wang}
(SBA) and, more recently, using Bayn-Kadanoff
functional theory (BKF) in terms of $X$-operators \cite{Zeyher,Zeyher1}.
On the basis of our Feynman path integral representation for the $t-J$
model \cite{Foussats1}, we developed a large-$N$ approach \cite{Foussats}
for the $J=0$ case
($U$-infinite Hubbard model).
This method has the advantage of working with Hubbard operators as
fundamental objects without any decoupling procedure.
Then, we neither take care of
gauge fluctuations nor Bose condensation as in the SBA.
The method developed in Ref.[\onlinecite{Foussats}]
was recently applied
to describe electronic
properties of quarter-filled organic molecular crystals \cite{Merino}.
For instance, in Ref.[\onlinecite{Merino}] we have
calculated electronic self-energies $\Sigma(k,\omega)$, spectral
functions $A(k,\omega)$ and the electronic density of states $N(\omega)$
which are nontrivial calculations considering that they involve
fluctuations, in a controllable approach, above the mean field.
Our $A(k,\omega)$ and $N(\omega)$ were carefully
compared \cite{Merino} with similar results
obtained by Lanczos diagonalization. Good agreement has been found for
the behavior of the dynamical properties at $J=0.0$.
Our previous studies for $J=0.0$ motivate us to
extend our approach to the case of finite $J$.
On the other hand, many physical problems play with a finite
value of $J$ and therefore,
it is necessary to extend our approach to this case.
This extension is the main topic of the present paper.
As our method is new and we claim it can be used to simplify several
calculations, we must show: a) how our method explicitly works
and, b) comparison of the results with others in the literature
in order to show the confidence of the method. To satisfy a) and b)
is one of the main purpose of the present paper.
In Sec.II we develop the perturbative expansion and we
give the new Feynman rules which are used in the explicit
calculation in Secs. III and IV.
In Sec.III we show results for the charge-charge correlation functions.
In Sec.IV we study different instabilities of the model.
In Sec. V we give the conclusions.
\section{Perturbative approximation: A large-$N$ approach}
\label{stripfil}
In this section, we present the large-$N$ expansion for the $t-J-V$ model
in the
framework of the path integral representation for Hubbard
operators. We will extend our formalism of Ref.[\onlinecite{Foussats}] to the case of
finite $J$.
As in Ref.[\onlinecite{Foussats}], our starting point for
developing the large-$N$ approach is the
path integral partition function written in the
Euclidean form ($it \rightarrow \tau$),
\begin{eqnarray}
Z&=&\int {\cal D}X_{i}^{\alpha \beta}\; \delta[X_{i}^{0 0} +
\sum_{\sigma} X_{i}^{\sigma \sigma}-1]\; \delta[X_{i}^{\sigma
\sigma'} - \frac{X_{i}^{\sigma 0} X_{i}^{0 \sigma'}}{X_{i}^{0 0}}]
\nonumber \\ &\times&(sdet M_{AB})_{i}^{\frac{1}{2}} exp\;(- \int
d\tau\;L_E(X, \dot{X}))\;.
\end{eqnarray}
In Eq.(1),
the five Hubbard $X$-operators $\hat{X}^{\sigma
\sigma'}$ and $\hat{X}^{0 0}$ are boson-like and the four Hubbard ${\hat
X}$-operators $\hat{X}^{\sigma 0}$ and $\hat{X}^{0 \sigma}$
are fermion-like
\cite{Hubbard}.
The spin index $\sigma$ is $\sigma = \pm$ ($up$ and $down$ state,
respectively).
There is also a superdeterminant
$(sdet M_{AB})^{\frac{1}{2}} = 1/\frac{1}{(-X_{i}^{0 0})^{2}}$
formed with the set of all the second class constrains of
the theory
(see Refs.[\onlinecite{Foussats}] and
[\onlinecite{Foussats1}]).
The Euclidean Lagrangian $L_E(X,\dot{X})$ in (1) is:
\begin{eqnarray}
L_E(X, \dot{X}) = \frac{1}{2} \sum_{i,
\sigma}\frac{({\dot{X_{i}}}^{0 \sigma}\;X_{i}^{\sigma 0} +
{\dot{X_{i}}}^{\sigma 0}\; X_{i}^{0 \sigma})}
{X_{i}^{0 0}}
+ H(X)\;.
\end{eqnarray}
On the basis of Hubbard $X$-operators,
the $t-J-V$ Hamiltonian is of the form:
\begin{eqnarray}
H(X) &=& \sum_{<ij>,\sigma}\;(t_{ij}\; \hat{X}_{i}^{\sigma 0}
\hat{X}_{j}^{0
\sigma} + h.c.) \nonumber \\
&+&
\frac{1}{2} \sum_{<ij>;\sigma} J_{ij}
(\hat{X}_{i}^{\sigma
{\bar{\sigma}}} \hat{X}_{j}^{{\bar{\sigma}} \sigma} -
\hat{X}_{i}^{\sigma
\sigma} \hat{X}_{j}^{\bar{\sigma} \bar{\sigma}})
\nonumber\\
&+& \sum_{<ij>;\sigma \sigma'} V_{ij} \hat{X}_{i}^{\sigma \sigma}
\hat{X}_{j}^{\sigma' \sigma'}
-\mu\sum_{i,\sigma}\;\hat{X}_{i}^{\sigma \sigma}.
\end{eqnarray}
Now, similarly to Ref.[\onlinecite{Foussats}]
we make the following changes in the path integral (1):
a) We
integrate over the boson variables ${X^{\sigma \sigma'}}$ using
the second ${\delta}$ -function in (1).
b) The spin index $\sigma=\pm$, is extended to a new index $p$
running from $1$ to $N$. In order to get a finite theory in the
$N$-infinite limit, we re-scale the hopping $t_{ij}$,
the exchange parameters
$J_{ij}$ and the nearest-neighbors Coulomb repulsion $V_{ij}$ to
$t_{ij}/N$, $J_{ij}/N$ and $V_{ij}/N$, respectively.
c) The completeness condition ($X_{i}^{0 0} + \sum_p
X_{i}^{pp}=N/2$) can be exponentiated, as usual, by using the
Lagrangian multipliers $\lambda_i$.
d) The charge-like terms (the 3rd and 4th term of (3)) of the Hamiltonian,
after extended to large-$N$, are written in terms of $X^{00}$ using
the completeness condition.
e)
We write $X^{00}$ and $\lambda$ in terms of static mean-field
values and dynamic fluctuations;
$X_{i}^{0 0} = N r_{0}(1 + \delta R_{i})$,
$\lambda_{i} = \lambda_{0} + \delta{\lambda_{i}}$.
f) Finally, we make the following change of variables;
$f^{+}_{i p} = \frac{1}{\sqrt{N r_o}} X_{i}^{p 0}$,
$f_{i p} = \frac{1}{\sqrt{N r_o}} X_{i}^{0 p}$.
By following the steps a-f, we find the effective Lagrangian
\begin{widetext}
\begin{eqnarray}
L_{eff}& = & - \frac{1}{2}\sum_{i,p}\left(\dot{f_{i p}}f^{+}_{i p}
+ \dot{f^{+}_{i p}}f_{i p}\right) \frac{1}{(1 + \delta R_{i})} +
\sum_{<ij>,p}(t_{ij}\ r_{o}f^{+}_{i p}f_{j p} + h.c.)
- (\mu - \lambda_{0})\;\sum_{i,p}\;f^{+}_{i p}f_{i p}
\frac{1}{(1 + \delta R_{i})} \nonumber \\
&+&
N\;r_{0}\;\sum_{i}\delta{\lambda_{i}}\;\delta R_{i}
+ \frac{1}{2}\sum_{<ij>,p,p'}\frac{J_{ij}}{N}\frac{f^{+}_{i p}
f_{i p'}}{ (1 + \delta R_{i})} \frac{f^{+}_{j p'} f_{i p}}{ (1 +
\delta R_{j})} + Nr_{0}^{2} \sum_{<ij>} (V_{ij}-
\frac{1}{2}J_{ij}) \delta R_{i}\delta R_{j} \nonumber\\
& +
&\sum_{i,p} f^{+}_{i p}f_{i p}\frac{1}{(1 + \delta R_{i})}\;
\delta{\lambda_{i}}+ L_{ghost}.
\end{eqnarray}
\end{widetext}
$(sdet M_{AB})^{\frac{1}{2}}$ leads to
$L_{ghost}({\bf {\cal Z}}) = - \sum_{i p}\; {\bf {\cal
Z}}^{\dag}_{i p}\left(\frac{1}{1 + \delta R_i}\right) {\bf {\cal
Z}}_{i p}$
when is written in terms of complex boson ghost field ${\bf {\cal Z}}_p$
\cite{Foussats}.
Now, we treat the exchange terms $J_{ij}$. These
can be decoupled in terms of the
bond variable $\Delta_{ij}$ through a Hubbard-Stratonovich transformation,
where $\Delta_{ij}$ is
the field associated with the quantity
$\sum_{p} \frac{f^{+}_{j p} f_{i p}}{ \sqrt{(1 + \delta R_{i})
(1 + \delta R_{j})}}$.
Finally, the Lagrangian (4) results
\begin{widetext}
\begin{eqnarray}
L_{eff}& = & - \frac{1}{2}\sum_{i,p}\left(\dot{f_{i p}}f^{+}_{i p}
+ \dot{f^{+}_{i p}}f_{i p}\right) \frac{1}{(1 + \delta R_{i})} +
\sum_{<ij>,p}\;(t_{ij}\ r_{o}f^{+}_{i p}f_{j p}+h.c.)
- (\mu - \lambda_{0})\;\sum_{i,p}\;f^{+}_{i p}f_{i p}
\frac{1}{(1 + \delta R_{i})}\nonumber \\
&+&
N\;r_{0}\;\sum_{i}\delta{\lambda_{i}}\;\delta R_{i}
+ \frac{2 N}{J}
\sum_{<ij>}\Delta^{+}_{ij}\;\Delta_{ij}- \sum_{<ij>}\left(\sum_{p}
\frac{f^{+}_{i p} f_{j p}}{ \sqrt{(1 +
\delta R_{i}) (1 + \delta R_{j})}}\; \Delta_{ij} + h.c.\right)\nonumber\\
& + & Nr_{0}^{2} \sum_{<ij>} (V_{ij}- \frac{1}{2}J_{ij})
\delta R_{i}\delta R_{j} + \sum_{i,p} f^{+}_{i p}f_{i
p}\frac{1}{(1 + \delta R_{i})}\; \delta{\lambda_{i}}
+ L_{ghost}.
\end{eqnarray}
\end{widetext}
At this point it is necessary to discuss similarities and differences between
the SBA and our approach.
a) From step f), we see that the fermions
$f_p$ are proportional to the
constrained $X^{0 p}$ operators.
They are not associated with the spinons
as in the SBA.
b) From step e), the field $\delta R$ is proportional to the
real $X^{00}$ operator representing the number of holes (empty sites).
This is not associated with the holons
as in the SBA.
c) The bond variable $\Delta_{ij}$ looks close to the
valence bond variable of the SBA.
However, $\Delta_{ij}$ besides to be a function of the correlated
fermions $f_p$, it is also
a function of $\delta R$ through the denominator.
Note that $L_{eff}$ (Eq.(5)) contains several nonpolynomial terms.
These apparent complications are the price we have to pay
for working in terms of
$X$-operators.
Since the $t-J-V$ model Hamiltonian is quadratic in the $X$'s,
the strong
electronic interactions are contained in the commutation rules and the
constraints.
In the path integral formulation, the information
contained in the Hubbard
algebra was transfered to the effective theory.
In the next sections we will show that present
theory can go beyond a
formal level and, the obtained results can be compared with others
in the literature.
Now, we write the $\Delta_{ij}$ fields in term of static mean-field
values and dynamics fluctuations
$ \Delta^{\eta}_{i}=\Delta (1+r^{\eta}_{i}+i A ^{\eta}_{i})$, where
${\eta}=x,y$ and $r^{\eta}_{i}$ and $A ^{\eta}_{i} $ correspond to the
amplitude and the phase fluctuations of the bond variable respectively.
To implement the $1/N$ expansion,
the nonpolynomial $L_{eff}$
should be developed, as in Ref.[\onlinecite{Foussats}],
in powers of $\delta R$. Up to order $1/N$
the following
Lagrangian is sufficient
\begin{widetext}
\begin{eqnarray}
&L_{eff}&=-\frac{1}{2}\sum_{i,p}\left(\dot{f_{i p}}f^{+}_{i p}
+ \dot{f^{+}_{i p}}f_{i p}\right) (1 - \delta R_{i} + \delta
R_{i}^{2})
+\sum_{<ij>,p}\;(t_{ij} r_{o} f^{+}_{i
p}f_{j p}+h.c.) - \mu \;\sum_{i,p}\;f^{+}_{i p}f_{i p} (1 - \delta R_{i}
+ \delta R_{i}^{2}) \nonumber \\ & + &
N\;r_{0}\;\sum_{i}\delta{\lambda_{i}}\;\delta R_{i} +\sum_{i,p}
f^{+}_{i p}f_{i p}(1 - \delta R_{i}) \;
\delta{\lambda_{i}} + \frac{2 N}{J}
\Delta^{2}\sum_{i\eta}\left[({r_{i}^{\eta}})^{2}+({A_{i}^{\eta}})^{2}\right]
\nonumber
\\
& -&\Delta \sum_{<ij>,p, p'} (f^{+}_{i p}f_{j p'}+f^{+}_{j p'}f_{i
p})[1 -\frac{1}{2} (\delta R_{i} + \delta
R_{j})+\frac{1}{4}\delta R_{i}\delta R_{j}+\frac{3}{8}(\delta
R_{i}^{2} + \delta R_{j}^{2})] \nonumber
\\ & -&\Delta \sum_{<ij>,p, p'} (f^{+}_{i p}f_{j p'}+f^{+}_{j
p'}f_{i p})({r_{i}^{\eta}}+i {A_{i}^{\eta}})[1 -\frac{1}{2}
(\delta R_{i} + \delta R_{j})] \nonumber
\\
& + & Nr_{0}^{2} \sum_{<ij>} (V_{ij}- \frac{1}{2}J_{ij})
\delta R_{i}\delta R_{j}
-\sum_{i p}\; {\bf {\cal Z}}_{i
p}^{\dag}\left(1- \delta R_i+ \delta R^2_i\right) {\bf {\cal
Z}}_{i p},
\end{eqnarray}
\end{widetext}
\noindent
where we have changed $\mu$ to $\mu-\lambda_0$ and dropped
constant and linear terms in the fields.
Looking at the effective Lagrangian (6), the Feynman rules can be
obtained as usual. The bilinear parts give rise to the propagators
and the remaining pieces are represented by vertices. Besides, we
assume the equation (6) written in the momentum space once the
Fourier transformation was performed.
In leading order of $1/N$, we
associate with the N-component fermion field $f_{p}$,
connecting
two generic components $p$ and $p'$,
the
propagator
\begin{eqnarray}
G_{(0)pp'}(k, \nu_{n}) = - \frac{\delta_{pp'}}{i\nu_{n} -
(E_{k} - \mu )}
\end{eqnarray}
\noindent
which is $O(1)$.
In (7),
$E_{k}=
-2(tr_{0}+\Delta) (cosk_x+cosk_y)$,
is the electronic dispersion in leading order, where
$t$ is the
hopping between nearest neighbors sites on the square lattice.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=7cm,angle=0]{./Feyman.eps}
\end{center}
\caption{Summary of the Feynman rules. Solid line represents the
propagator $G_{(0)}$ (Eq.(7)) for the correlated fermion $f_p$. Dashed line
represents the $6 \times 6$ boson propagator
$D_{(0)}$ (Eq.(8))
for the $6$-component field $\delta X^a$. Note that the component
$(1,1)$ of this propagator is directly associated
with the $X^{00}$ charge operator. Doted line is the propagator
$B$ for the boson ghost field ${\cal Z}_p$.
$\Lambda^{pp'}_a$ (Eq.(9)) and $\Lambda^{pp'}_{ab}$ represent
the interaction between two fermions $f_p$ and one and two bosons
$\delta X^a$ respectively.
$\Gamma^{pp'}_a$ and $\Gamma^{pp'}_{ab}$ represent
the interaction between two ghost fields ${\cal Z}_p$ and one and two bosons
$\delta X^a$ respectively.
}
\label{}
\end{figure}
The mean field values $r_0$ and $\Delta$ must be determined
minimizing the leading order theory.
From the completeness condition $r_0$ is
equal to $\delta/2$ where $\delta$ is the hole doping away from
half-filling. On the other hand, minimizing respect to $\Delta$
we obtain $\Delta=\frac{J}{2} \frac{1}{N_s}
\sum_k cos(k_x) n_F(E_k-\mu)$ where $n_F$ is the Fermi function and
$N_s$ is the number of sites in the Brillouin zone (BZ).
For a given doping $\delta$,
the chemical potential $\mu$ and $\Delta$
must be determined self-consistently from
$(1-\delta)=\frac{2}{N_s} \sum_{k} n_F(E_k-\mu)$.
We associate with the six component $\delta X^{a}
= (\delta R\;,\;\delta{\lambda},\; r^{x},\;r^{y},\; A^{x},\;
A^{y})$ boson field, the inverse of the propagator,
connecting two generic components a and b,
\begin{widetext}
\begin{eqnarray}
D^{-1}_{(0) ab}(q,\omega_{n})= N \left(
\begin{array}{cccccc}
(4V-2J)r_{0}^{2}(cos(q_{x})+cos(q_{y}))
& r_{0} & 0 & 0 & 0 & 0 \\
r_{0} & 0 & 0 & 0 & 0 & 0 \\
0 & 0 & \frac{4}{J}\Delta^{2} & 0 & 0 & 0 \\
0 & 0 & 0 & \frac{4}{J}\Delta^{2} & 0 & 0 \\
0 & 0 & 0 & 0 & \frac{4}{J}\Delta^{2} & 0 \\
0 & 0 & 0 & 0 & 0 & \frac{4}{J}\Delta^{2} \
\end{array}
\right)
\end{eqnarray}
\end{widetext}
\noindent
The bare boson propagator $D_{(0)ab}$
(the inverse of Eq.(8)) is $O(1/N)$.
As we will see
$D_0$ is renormalized to $D$ by
an infinite series of diagrams of $O(1/N)$.
We associate with the N-component ghost field ${\bf {\cal
Z}}_{p}$, the propagator,
connecting two generic components $p$ and $p'$,
$B_{pp'} = - \delta_{pp'}$
which is $O(1)$.
\begin{widetext}
The expressions for three-leg and
four-leg vertices are:
a)
\begin{eqnarray}
\Lambda^{pp'}_{a}& =& (-1)\; \left(\frac{i}{2}(\nu_n + {\nu'}_n) +
\mu + 2\Delta \sum_{\eta} cos(k_\eta-\frac{q_\eta}{2})\;
cos\frac{q_\eta}{2};\;1;
- 2\;\Delta\; cos(k_x-\frac{q_x}{2}) \; ; \right.\nonumber \\
&-& \left. 2\;\Delta\;
cos(k_y-\frac{q_y}{2}) ;\right.
\left. 2\; \Delta\; sin(k_x-\frac{q_x}{2}) \;;2 \Delta\;
sin(k_y-\frac{q_y}{2}) \right)\;\delta^{pp'}
\end{eqnarray}
\noindent
represents the interaction between two fermions and
one boson.
b)
$\Lambda^{pp'}_{ab}$,
which represents the interaction between two fermions and
two bosons, is a $6 \times 6$ matrix where the only elements different
from zero are:
\begin{eqnarray}
\Lambda^{pp'}_{\delta R \delta R}& =& \left(\frac{i}{2} (\nu_n + {\nu'}_n)
+ \mu
+ \Delta\; \sum_{\eta}
cos(k_\eta-\frac{q_\eta+q'_\eta}{2})\;
[ cos\frac{q_\eta}{2} \; cos\frac{q'_\eta}{2}\;
+\;cos\frac{q_\eta+q'_\eta}{2}] \right)
\delta^{pp'}
\end{eqnarray}
\end{widetext}
\begin{eqnarray}
\Lambda^{pp'}_{\delta R \delta\lambda}=\frac{1}{2}
\;\delta^{pp'}
\end{eqnarray}
\begin{eqnarray}
\Lambda^{pp'}_{\delta R \; r^{\eta}}= -\Delta \;
cos(k_\eta-\frac{q_\eta+q'_\eta}{2})\; cos\frac{q'_\eta}{2}
\;\delta^{pp'}
\end{eqnarray}
\begin{eqnarray}
\Lambda^{pp'}_{\delta R \; A ^{\eta}}= \Delta \;
sin( k_\eta-\frac{q_\eta+q'_\eta}{2})\; cos\frac{q'_\eta}{2}
\;\delta^{pp'}
\end{eqnarray}
c)
$\Gamma^{pp'}_{a} = (-1) (\delta_{pp'}\;,\;0,\;0,\;0,\;0,\;0)$
represents the interaction between two ghosts and one
boson.
d) ${\Gamma}^{pp'}_{ab}$ is a $6\times6$ matrix,
where ${\Gamma}^{pp'}_{{\delta} R {\delta} R}=\delta_{pp'}$ and,
the others components are zero.
It represents the interaction between two bosons and two
ghosts.
Each vertex conserves the momentum and energy and they are $O(1)$.
In addition, in each diagram there is a minus sign for each fermion loop and
a topological factor.
Fig.1 summarizes the Feynman rules.
After identifying the propagators and vertices and the respective
order of them, in the next sections, we will
calculate different physical quantities.
Before finishing this section, one remark is necessary. From the $N$-extended
completeness condition we can see that the charge operator
$X^{00}$ is $O(N)$, while the operators $X^{pp}$ are $O(1)$.
This fact will have the physical consequence that the $1/N$ approach
weakens the effective spin interactions compared to the one related to the
charge degrees of freedom. This is discussed in
the next section.
\section{Charge Correlations}
In this section density-density correlations functions are calculated.
It will be showed that the developed
formalism can be used in explicit
calculation of different correlation functions and
we will also compare our results with others in the literature.
The density-density correlation function is defined as \cite{Lew,Foussats}
\begin{eqnarray}
\tilde{D}_{ij}=\frac{1}{N} \sum_{pp'}
<T_{\tau} X^{pp}_i X^{p'p'}_j>
\end{eqnarray}
Using $\sum_p X^{pp}_i=N/2- X^{00}_i$ we find for $\tilde{D}$
in
the Fourier space
\begin{eqnarray}
\tilde{D}({\bf q},\omega_n)=
-N {(\frac{\delta}{2})}^2 D_{\delta R \delta R}({\bf q},\omega_n)
\end{eqnarray}
The charge correlation, in $O(1)$, needs the calculation of
all $O(1/N)$ contributions to
$D_{\delta R\delta R}({\bf q},\omega_n)$.
From the Dyson equation, $(D_{ab})^{-1} = (D_{(0) ab})^{-1} - \Pi_{ab}$,
the dressed components $D_{ab}$ of
the boson propagator can be found after
the evaluation of the $6 \times 6$ boson self-energy matrix $\Pi_{ab}$.
Using the Feynman rules we may evaluate
$\Pi_{ab}$ through the diagramas of Fig.2.
Note that $D_{\delta R \delta R}$ is the element $(1,1)$ of
the $6\times6$
dressed propagator.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=7cm,angle=0]{./Irreducible.eps}
\end{center}
\caption{The four different contributions
$\Pi^{(i)}_{ab}$ ($i=1,2,3,4$) to
the irreducible boson self-energy $\Pi_{ab}$.
}
\label{}
\end{figure}
We note here an important difference with respect to the SBA.
In SBA, only in leading order, the charge-charge
correlation can be associated with the holon propagator.
Beyond leading order, the convolution of
holon propagators (which means the
reconstruction of the $X$-operator $X^{00}$)
is necessary.
Meanwhile, in our case, it is not used
any decoupling scheme and then, the different
correlation functions are
directly associated with our field variables.
The density-density spectral function,
which in principle can be measure with
electron energy loss scattering \cite{EELS},
is defined as
$D({\bf q},\omega)=-Im[\tilde{D}({\bf q},\omega)]$.
The imaginary part is taken as usual after performing the
analytical
continuation $i\omega_n=\omega+i\eta$.
In Fig.3 we show the density-density correlation function for the
physical
value $J=0.3$ and for doping $\delta=0.20$.
We plotted the densities for different ${\bf q}$ vectors in the
BZ as a function of $\omega$.
The ${\bf q}$ vectors and the physical parameters are
the same to those used in the calculation of the densities
in Ref.[\onlinecite{Zeyher2}]. We used $\eta=0.1$ in the
analytical continuation.
Comparing Fig.3 with Fig.3 of Ref.[\onlinecite{Zeyher2}],
we find a remarkable agreement between both methods in
spite of the fact that
the two
approaches are very different.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=8cm,angle=0]{./charge.eps}
\end{center}
\caption{Density-density correlation functions for $J=0.3$, $V=0.0$ and
$\delta=0.20$ for different ${\bf q}$'s in the BZ.
The density-density correlation functions contain collective peaks
due to infinite
diagram summation seen in Fig.2. The collective peaks
are well pronounced when they are above
the particle-hole continuum (see for example ${\bf q}=(\pi,0)$).
They are less pronounced when they are superimposed
to the particle-hole continuum (see for example ${\bf q}=(2\pi/5,\pi/5)$).
$\omega$ is in units of $2t$ and $t$ is considered to be $1$.
}
\label{}
\end{figure}
The density-density
correlation function is nearly independent of $J$.
For example, comparing Fig.3 (for J=0.3) with Fig.2 in
Ref.[\onlinecite{Foussats}](for $J=0$) for momentum
${\bf q}=(\pi,\pi)$, we find a well
pronounced collective peak (or zero sound)
at $\omega \sim 3.5$.
(Note that in Ref.[\onlinecite{Foussats}]
$\omega$ is in units of $t$ meanwhile in Fig. 3 it is in units
of $2t$ in order to absorb the factor $1/N$ due
to the re-scaling of the
hopping term.)
The fact that the density correlations are nearly independent of $J$
means that $\tilde{D}({\bf q},i \omega_n)$ is
dominated by charge fluctuations.
As we mentioned above, the present formalism privileges charge over
spin fluctuations.
Another consequence of this result is the absence, in $O(1)$,
of collective excitations (like magnons) in the spin
susceptibility.
The spin-spin correlation function is
the electronic bubble with renormalized band due to
correlations \cite{Foussats,Lew}.
Meanwhile there are collective effects in the charge sector in $O(1)$,
that appear
in $O(1/N)$
in the spin sector.
The collective peaks are more pronounced when
they are present
well above the particle-hole continuum as for ${\bf q}=(\pi,\pi)$.
For momentum
${\bf q}=(2 \pi/5, \pi/5)$ the collective peak is superimposed to
the particle-hole
continuum and it appears broader.
The broadening of the collective peak ${\bf q}=(\pi,\pi)$
is not intrinsic and it is
only due to the finite value of $\eta$ we used in the analytical continuation.
At this point we can compare
our results with those obtained using exact
diagonalization \cite{Tohyama}.
Despite the charge correlation function in Ref.[\onlinecite{Tohyama}]
was for doping $\delta=0.25$
and $J=0.4$
and our calculation is for $\delta=0.20$ and $J=0.3$,
there are some similarities and differences to remark.
For example, Fig. 2 in Ref.[\onlinecite{Tohyama}]
shows also a peak at ${\bf q}=(\pi,\pi)$.
However, a) this peak
appears at larger energies than our collective peak,
b) the peak at $(\pi,\pi)$ obtained
in Ref.[\onlinecite{Tohyama}] contains
more structure than the ours and c) we must note also a difference at
${\bf q}=(\pi,0)$.
In our calculation we found a collective-like peak at
$\omega \sim 2$ and
a broad and small continuum
at low energy.
Lanczos diagonalization presents also this picture but,
in addition there is a peak at lower energy.
As it was already pointed out by Khaliullin and Horsch
in Ref.[\onlinecite{Kha1}] (see also Ref.[\onlinecite{Kha2}]),
the inclusion
of fluctuations beyond the mean field level is probably responsible
for the differences listed in a), b) and, c).
We think that the inclusion of fluctuations is more important at lower
than at larger doping.
In Ref.[\onlinecite{Merino}], for doping
$\delta=0.5$, we found a better agreement between
the peak position for ${\bf q}=(\pi,\pi)$ obtained by our approach
and that obtained by Lanczos.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=7cm,angle=-90.]{./Jinstvsdo.eps}
\end{center}
\caption{Phase diagram in the $\delta_c-J$ plane for $V=0.0$.
The solid line defines the stability border for the homogeneous
Fermi liquid (HFL). For a given $J$ the HFL is stable for $\delta>\delta_c$.
The thick dashed line around $J \sim 0.5$ separates the flux
phase (FP) from the bond order phase
(BOP).
}
\label{}
\end{figure}
\section{Instabilities}
\subsection{Flux and Bond-Order phases}
The theory developed in previous sections defines a
homogeneous Fermi liquid
(HFL)
phase.
In $O(1)$ the mean field solution ($r_0$, $\Delta$) was independent of
sites. Therefore, at $N$-infinite,
we have free fermions with renormalized
band $E_k$ due to correlations.
In this section we will study the stability conditions for the HFL.
In leading order, there are collective effects in the charge
sector whereas there are not in the spin sector.
Therefore, in principle, we expect
instabilities in the charge channel.
The HFL system is unstable when
the static charge susceptibility
$Re(\tilde{D}({\bf q},i\omega_n=0)$
diverges. The ${\bf q}$ vector,
where the instability occurs ($\bf{q_c}$), being the modulation
of the new phase.
Fig.4 shows the phase diagram of the model for $V=0$.
For a given $J$, below a critical doping $\delta_c$
where the static charge
susceptibility diverges at a given vector ${\bf q}$ in the BZ,
the HFL is not the stable phase.
In the limit $J \rightarrow 0$
$\delta_c \rightarrow 0$ the HFL is stable for
the whole doping range except
at half filling
where the system is an insulator because the band effective-mass
tends to infinite.
The instability is placed, for all $J$, on the border of the BZ.
That is, ${\bf q_c}=(1,x) \pi$
or ${\bf q_c}=(x,1)\pi$
where
the parameter $x$ measures the degree of the
incommensuration of the instability.
In Fig.5 we plot the incommensuration $x$ as a function of $\delta_c$.
For $\delta_c \rightarrow 0$, $x \rightarrow 1$. Therefore, in
the limit of $J \rightarrow 0 $, we found a phase
with commensurate
order ${\bf q_c}=(\pi,\pi)$.
It is easy to overlook the instability just by looking at the
static charge susceptibility.
In Fig.6 we plot the static charge susceptibility, along
${\bf q}=(\pi,q)$, for
$\delta=0.15$ and $\delta=0.14$ and, for the physical value $J=0.3$.
From Fig.4 we know that the HFL is not stable for doping $\delta=0.14$
whereas for $\delta=0.15$ it is stable.
Both curves in Fig.6 look similar and, for $\delta=0.14$
there is no indication of the instability for ${\bf q} \sim {\bf q_c}$
because it occurs
in a very narrow region of momentum near ${\bf q_c}$.
In practice,
this means that the
set of ${\bf q}$-points in Fig.6 is
not dense enough to localize the divergence of
$Re(\tilde{D}({\bf q},i\omega_n=0))$.
As we will see below, this is related with the fact that the instability
is weakly coupled with the charge sector.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=8cm,angle=0]{./inconmvsdop.eps}
\end{center}
\caption{Inconmensuration $x$ vs $\delta_c$ for $V=0.0$.
$x$ is defined as ${\bf q_c}=(1,x)\pi$.
${\bf q_c}$ is the ${\bf q}$-vector
where the instability takes place.
The instability is incommensurate for all $J$ except
for $J \rightarrow 0$
where $x \rightarrow 0$ (${\bf q_c} \rightarrow
(\pi,\pi)$). At the separating border of the FP from the BOP there is a
jump in the value of ${\bf q_c}$.
}
\label{}
\end{figure}
To calculate $\tilde{D}({\bf q},\omega_n)$
we have to evaluate the inverse
of $(D_{ab})^{-1}$.
Therefore, for a better determination of the instability,
we may look for the zeros of the
determinant of $D_{ab}^{-1}$.
In the inset of Fig.6 we plot the $det(D^{-1}_{ab})$ as a function of
${\bf q}=(\pi,q)$. For $\delta=0.14$
(close to the onset of the instability)
the determinant changes
sign in a narrow region around ${\bf q_c} \sim (\pi,2.7)$ making the HFL
unstable. Meanwhile, for
$\delta=0.15$ the determinant is positive for all {\bf q}
and the system is stable.
A similar plot for $\delta < 0.14$ shows,
of course, a larger negative region
for $det(D_{ab}^{-1})$.
In order to characterize the nature of the new phase, we have studied
the eigenvector corresponding to the zero eigenvalue of
$(D_{ab})^{-1}$.
For doping $\delta < \delta_c \sim 0.5$
the eigenvector corresponding to the zero mode
is $\sim (0,0,0,0,-1,1)$. Then, in the new phase,
the 5th and 6th components
of $\delta X^a$ ($A^x$ and $A^y$) are frozen.
$A^x$ and $A^y$ are associated with the phase
of the field $\Delta^{\eta}$
and they take opposite values.
Therefore, as the instability occurs near the momentum $(\pi,\pi)$,
this new phase is the well known flux phase (FP)
\cite{Marston,Capellutti}
which opens a gap,
with $d$-wave symmetry
in the normal state.
Recently,
Chakravarty, Laughlin, Morr and Nayak
considered that the FP
or DDW as a candidate to explain the physics
of the pseudogap
in underdoped cuprates\cite{Chakra}.
The first component of the eigenvector corresponding
to zero eigenvalue
is not zero exactly. It has a small value which
means that the FP is weakly
coupled to the charge sector.
It means that the DDW is not completely hidden when it is
incommensurate. Although Bragg peaks
are predicted, they will show low intensity which would make their
observation very difficult.
This is the reason by which,
any charge-probe is not very sensitive
to show the FP instability.
When ${\bf q_c}$ is exactly $(\pi,\pi)$
the eigenvector does not have any
mixing with the charge sector and, the flux phase is fully hidden.
For doping $\delta > \delta_c \sim 0.5$ the unstable
eigenvector is of the form
$\sim (0,0,0,1,0,0)$. This state is known as bond-order phase (BOP)
\cite{Capellutti,
Lubenski}.
In the range of the studied parameters we did not find indications
for phase separation.
In Fig.5,
for $J \sim 0.5$, near the crossover from the
FP to BOP, there is a jump in the
incommensuration $x$.
The agreement with other methods
\cite {Capellutti,Lubenski}
is again remarkable in spite of the fact that the
present approach is, {\it a priori}, very different from those ones.
Taking into account that we constantly find
similar or even the same results
by means of different methods turn the results reliable.
For example,
for the physical value $J=0.3$, the HFL is unstable against
a FP.
This result is very robust and, many different
methods agree that the stable
phase, for small doping, is a FP for the physical
region of the parameters
\cite{Marston,Grilli,Lubenski,Capellutti,
Numerics}.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=8cm,angle=0.]{./inset.eps}
\end{center}
\caption{Static charge susceptibility
$Re(\tilde{D}({\bf q},i\omega_n=0))$
vs ${\bf q}=(\pi,q)$ for
$J=0.3$, $V=0.0$ and, for $\delta=0.15$ and $\delta=0.14$.
For $\delta=0.14$ (solid line) the curve looks similar
to the case for $\delta=0.15$ (dashed line). For $\delta=0.14$ the HFL
is not stable and there is
no strong sign of the instability in the static charge susceptibility.
In the inset we plot the $det (D_{ab}^{-1})$ (see text for discussions)
vs ${\bf q}=(\pi,q)$. For $\delta=0.14$, the $det (D_{ab}^{-1})$ clearly
changes sign
in a small region around ${\bf q_c} \sim (\pi,2.7)$.
}
\label{}
\end{figure}
\subsection{Charge Density Wave phase}
The inclusion of a nearest-neighbors Coulomb repulsion $V$
favors a charge density wave (CDW) state.
In order to investigate this instability,
we have included a finite value of $V$ in the calculation.
The study of a nearest-neighbors Coulomb repulsion
in correlated models is important.
In cuprates, there are indications \cite{Fein} showing the existence
of $V$ with a value close to $J$.
If $J$ favors superconductivity and, the value of $V$ is large,
it is reasonable to think
that superconductivity will be diminished due the
presence of $V$.
There are some analytical and numerical works
studying the competition between $V$ and $J$ on superconductivity.
Meanwhile some papers indicate that superconductivity in the $t-J$ model
vanishes for $V \sim J$ \cite{Zeyher1}, others indicate that
superconductivity survives
up to values of $V >> J$ \cite{Riera}.
On the other hand, the presence of $V$ seems to be important
for understanding the physics of organic materials \cite{jaimes}.
Some peculiarities in the optical conductivity
(and also superconductivity)
occur near the charge order and this
picture can be interpreted by
the competition between $V$ and the kinetic energy in correlated
models \cite{org}.
Looking at the Eq.(8) we see that $V$ is only present
in the element $D^{-1}_{(0)}(1,1)$.
The Coulomb term enters in our approach
multiplied by $(\delta/2)^2$ which means that the effect of $V$,
at low doping,
is strongly
screened by
the correlations.
We note that in the SBA $V$ can enter in different manners
depending on the way chosen for the decoupling.
In Fig.7 we show the phase diagram in the $V_c-\delta$ plane
for J=0 and $J=0.3$.
The curves
show, as a function of $\delta$, the
critical Coulomb repulsion $V_c$ where the CDW instability takes place.
For $V <V_c$ the HFL is stable.
This result is in agreement with numerical \cite{Merino} and
analytical methods
\cite{Hoang}.
In Ref.[\onlinecite{Hoang}] the authors use coherent
potential approximation and
their results are
close to the ours.
A similar result was also recently found, at $J=0$, in the
triangular lattice in the context of the new
low dimensional superconductor
$Na_xCoO_2$ \cite{Lee}.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=7cm,angle=-90.0]{./VCvsdopJ0.eps}
\end{center}
\caption{Charge density wave (CDW) stability line $V_c$ vs $\delta$
for $J=0.0$ (dashed line) and $J=0.3$ (solid line). For a given $\delta$,
the system is in a CDW state for $V > V_c$. The value of $V_c$ increases
with increasing $J$.
For $J=0.3$ the first presented doping is $\delta=0.15$.
(For $J=0.3$ the HFL is not stable for $\delta <0.14$
at $V=0.0$).
}
\label{}
\end{figure}
For $J=0.3$ and doping $\delta=0.20$,
the HFL is stable at $V=0$.
When $V= V_c\sim 1.25$
the system enters in a CDW state. This
$V_c$ is $\sim$ $40 \%$
larger than for $J=0$
and the same doping.
There are two sources
for
this tendency.
a) When $J$ is finite, there is a contribution $\Delta$ to the effective
hopping in $E_k$. Then, the system win kinetic energy when $J$
is finite and, a larger $V$ is necessary
in order to localize the charges.
b) As we see in the element $(1,1)$ of $D^{-1}_{(0)}$, the effect of $V$
is diminished
when $J$ is finite.
The term $2J$ in the element $(1,1)$ comes
from the charge-like term
$J_{ij}X_{i}^{\sigma
\sigma} X_{j}^{\bar{\sigma} \bar{\sigma}}$
of the pure $t-J$ model, which is
of the same form of the Coulomb term (see Eq.3).
In order to identify the main source for the increasing of $V_c$
we looked for
the CDW instability without the charge-like term in (3).
Under this condition, for $J=0.3$ and $\delta=0.20$,
$V_c$ is $V_c \sim 1.0$. This value is nearly the
same than $V_c$ for $J=0$. Therefore, b) is the main source
for the increasing of $V_c$ with increasing $J$.
In contrast to the FP and BOP,
the CDW instability can be detected clearly
by looking at the static charge susceptibility
$Re(\chi({\bf q},i\omega_n=0))$.
In Fig.8 we
show,
for $J=0.3$ and $\delta=0.25$,
$Re(\chi({\bf q},i\omega_n=0))$ vs ${\bf q}=(q,q)$
for $V=1.$ (a
little smaller than the critical value $1.1$) and $V=0.5$.
In the figure, we see that approaching $V_c$, the static
susceptibility tends to diverge near ${\bf q}=(\pi,\pi)$.
For $V=0.5$ ($V<<V_c$), the static charge susceptibility is flat.
\begin{figure}
\begin{center}
\setlength{\unitlength}{1cm}
\includegraphics[width=7cm,angle=-90.0]{./CharvskV1.eps}
\end{center}
\caption{Static charge susceptibility for $J=0.3$ and $\delta=0.25$ vs
${\bf q}=(q,q)$ for two different values of $V$, $V=1.$ and $V=0.5$.
$V=1.$ is near (from below) the value of $V_c=1.1$.
In contrast to Fig.6, the CDW instability has a clear signal in the
static charge susceptibility. For $V=1.$ there is a clear
indication of a divergence at $(\pi,\pi)$.
For $V=0.5$ ($V<<V_c$), the static charge susceptibility is flat.
}
\label{}
\end{figure}
In this case the eigenvector corresponding to the zero eigenvalue is
$ \sim (1,0,0,0,0,0)$.
Then, the instability is mainly in the pure charge sector.
From Fig.8, in agreement with other methods \cite{Merino},
the instability
${\bf q}$-vector is always at ${\bf q}=(\pi,\pi)$.
Therefore, in the CDW phase, the charges order themselves
forming a checkerboard pattern.
For doping different to the commensurate one $\delta=0.5$ we must
interpret the CDW instability as the system is charge-ordered
but remains metallic.
These results were obtained by different methods and the
nature of the instability for doping different to the commensurate
one is still open.
For the commensurate doping $\delta=0.5$ in Ref.[\onlinecite{Merino}]
we showed that the system is
charge-ordered at $V_c \sim 0.6$ however, we also find
an insulator because the quasiparticle
weight goes to zero exactly at this $V_c$ value.
In order to compare the above results to
the predictions of the BKF method, we calculated
the critical $V_c$ using the formulation of Refs.[\onlinecite{Zeyher}]
[\onlinecite{Zeyher1}] for $J=0.3$.
For $\delta=0.20$ and $\delta=0.60$ we found
$V_c\sim 1.25$ and $V_c\sim 0.9$
respectively. The agreement with our calculation is again very good
in spite of the fact that
the Coulomb term enters BKF
in a different way.
\section{Conclusions}
From the path integral representation
for the Hubbard
$X$-operators we have developed a new perturbative theory for the $t-J-V$
model.
Our formulation is free of slave-particles
and the $X$-operators are treated as fundamental objects.
We have extended the large-$N$ formalism, recently developed in Ref.
[\onlinecite{Foussats}], to the case of finite $J$.
To work directly with the $X$-operators has several advantages upon the
slave-particles methods.
For example, in the slave-boson theory,
the decoupling procedure introduces a
gauge degree of freedom and then, an additional problem as the gauge
fixing arises.
This problem does not occur in our formulation.
On the other hand, in the slave approaches, in contrast to our method,
beyond a mean field level, a convolution between
slave-particles propagators
is necessary to reconstruct the original
physical propagator.
Fig.1 summarizes the Feynman rules of the present approach
where the propagators and vertices are written in terms of the
Hubbard operators $X^{\sigma 0}$, $X^{00}$, etc.
It is important to note that
we can not read the interaction vertices from the
$t$ and $J$ terms of the Hamiltonian.
They arise from the nonpolynomial effective theory defined by $L_{eff}$
in Eq.(6).
All the interactions caused by the
Hubbard algebra (commutation rules and constraints) were transfered to
$L_{eff}$.
The formalism developed in the present paper privileges
charge over spin fluctuations.
While there are collective effects in the charge sector in $O(1)$,
those appear in $O(1/N)$ in the spin sector.
We have studied
charge correlations functions and we have investigated the role of
collective effects.
This study shows the presence of collective peaks (or zero sound)
which are more evident when they are separated from the
particle-hole continuum. On the other hand, the collective peaks become
broader when they are superimposed to the particle-hole continuum.
One important characteristic of the charge correlation function is the
fact that this is not strongly dependent on the value of the
exchange interaction $J$.
At large $N$, the theory can be described as a homogeneous Fermi liquid
with renormalized band due to correlations.
We have studied the stability of this Fermi liquid phase.
For a given value of $J$, the Fermi liquid phase is stable for doping
$\delta > \delta_c$. The critical doping $\delta_c$ decreases with
decreasing $J$.
For doping below $\delta_c$ the system enters in a flux or
bond order phase
depending if $J < 0.5$ or $ J > 0.5$ respectively.
These instabilities are
weakly coupled to the charge sector which means that any charge-probe
is not an efficient test to detect them.
One important characteristic of these new phases is that they are
incommensurate with a modulation vector ${\bf q}=(1,x)\pi$ where the
incommensuration $x$ tends to $1$ when $J$ goes to zero.
It is important to remark that in agreement with other theories,
for low doping, the Fermi liquid is unstable
against a flux or $d$-density wave phase for the physical region of
the parameters.
We also have investigated the role of a nearest-neighbors
Coulomb repulsion
$V$ on the stability conditions of the Fermi liquid.
When $V$ is larger than a critical value $V_c$, the system enters,
at a given doping, in charge density wave
state. The value of $V_c$ increases
with increasing
$J$. We have identified in the charge-like term of the pure $t-J$ model
the main reason for this increase.
We have continuously compared our results with similar ones
in the literature.
The agreement of our results with those obtained by
other methods give
confidence to the results and the approach of the present paper.
At leading order, our formalism is in agreement with the slave-boson
approach. However, at the next to leading order
(which is necessary to calculate dynamical properties)
the differences between the two
formulations are not yet completely established.
We think that the complications that appear in
the slave-boson method beyond mean field level
as gauge fixing, Bose condensation,
regularizing factors \cite{Arrigoni} are not good
for the advance of the field.
In our case we do not have these problems and we think that our approach
can be useful to go beyond the mean field level.
With the formulation for finite $J$ in hand,
we expect to continue in this direction as we did
for $J=0$ in Ref.[\onlinecite{Merino}].
\noindent{\bf Acknowledgments}
The authors thank to C. Genz,
L. Manuel, J. Merino and R. Zeyher
for valuable discussions.
%\begin{references}
\begin{thebibliography}{99}
\bibitem{Anderson} P.W. Anderson, The Theory of Superconductivity in
High-$T_c$ Cuprates (Princeton University Press, Princeton,1997).
\bibitem{McKenzie} R.H.McKenzie, Science {\bf 278},820(1997).
\bibitem{Izyumov} A. Izyumov, Physics-Uspekhi {\bf 40}, 445 (1997).
\bibitem{Dagotto} E. Dagotto, Rev.Mod.Phys. {\bf 66},763(1994).
\bibitem{Hubbard} J. Hubbard, Proc. R. Soc. London Ser. A {\bf 276},
238 (1963).
\bibitem{Arrigoni} E. Arrigoni et al, Phys. Rep.{\bf 241}, 291(1994).
\bibitem{Wang} Z. Wang, Int. Journal of Modern Physiscs B {\bf 6}, 155(1992).
\bibitem{Zeyher} R. Zeyher and M. L. Kuli\'c, Phys.Rev. B {\bf 53}, 2850(1996).
\bibitem{Zeyher1} R. Zeyher and A. Greco, Eur. Phys. J.B {\bf 6}, 473 (1998).
\bibitem{Foussats1} A. Foussats, A. Greco, C. Repetto, O. P. Zandron and O. S.
Zandron, Journal of Physics A {\bf 33}, 5849(2000).
\bibitem{Foussats} A. Foussats and A. Greco; Phys.Rev.B{\bf 65},
195107(2002).
\bibitem{Merino} J. Merino, A. Greco, R. H. McKenzie, and M. Calandra,
Phys. Rev. B {\bf 68}, 245121(2003).
\bibitem{Lew} L. Gehlhoff and R. Zeyher,
Phys. Rev. B{\bf 52}, 4635(1995).
\bibitem{EELS} E. Zojer, M. Knupfer, Z. Shuai, J. Fink, J.L. Bredas,
H. H. Horhold, J. Grimme, U. Scherf, T. Benincori, and G. Leising,
Phys.Rev.B {\bf 61},16561(2000).
\bibitem{Zeyher2} R. Zeyher and M. L. Kuli\'c, Phys.Rev. B {\bf 54}, 8985(1996).
\bibitem{Tohyama} T. Tohyama, P. Horsch and S. Maekawa, Phys. Rev. Lett.
{\bf 74}, 980(1995).
\bibitem{Kha1} G. Khaliullin and P. Horsch; Phys.Rev. B {\bf 54}, R9600(1996).
\bibitem{Kha2} P. Horsch and G. Khaliullin; cond-mat/0312561 (2003).
\bibitem{Marston} I. Affleck and J.B. Marston, Phys.Rev.B{\bf37},3774(1988).
\bibitem{Capellutti} E. Cappelluti and R. Zeyher, Phys.Rev.B{\bf 59},
6475(1999).
\bibitem{Chakra} S. Chakravarty, R. B. Laughlin, D.K.Morr, and Ch. Nayak, Phys.
Rev. B {\bf 63},94503 (2001).
\bibitem{Grilli} M. Grilli and G. Kotliar, Phys.Rev. Lett. {\bf 64}, 1170
(1990).
\bibitem{Lubenski} D.C. Morse and T.C. Lubensky,
Phys.Rev. B {\bf 42},7994(1990).
\bibitem{Numerics} P.W.Leung, Phys. Rev. B{\bf 62},6112(2000).
\bibitem{Fein} L.F. Feiner,J.H.Jefferson and R. Raimondi, Phys.Rev. B {\bf 53}
8751 (1996).
\bibitem{Riera} J. Riera and E. Dagotto, Phys.Rev.B {\bf 57},8609 (1998);
M. Calandra and S. Sorella, Phys. Rev. B{\bf 61},11894(2000); C. Gazza,
G.B. Martins, J. Riera and E. Dagotto, Phys.Rev.B {\bf 59}, 709(1999).
\bibitem{jaimes} R.H.McKenzie,J.Merino,J.B.Marston and O.P.Sushkov,
Phys.Rev.B{\bf 64},085109(2001); J.Merino and R.H.McKenzie, Phys.Rev. Lett.
{\bf 87},237002(2001).
\bibitem{org} M.Dressel, N.Drichko, J.Schlueter, and J. Merino,
Phys.Rev.Lett.{\bf 90},167002(2003).
\bibitem{Hoang} A.T. Hoang and P.Thalmeier, J. Phys.: Condens. Matter
{\bf 14}
6639(2002).
\bibitem{Lee} O. Motrunich and P. Lee, Phys. Rev.B{\bf70},024514(2004).
\end{thebibliography}
%\end{references}
\end{document}