Mesoscopic BCS pairing in the repulsive 1d-Hubbard model
Extracted source text ยท arXiv cond-mat/0509554
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%% DOCUMENT %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\documentclass[preprint,amssymb,amsmath,aps,prb,showpacs]{revtex4}
\usepackage{graphicx}
\begin{document}
\title{Mesoscopic BCS pairing in the repulsive 1d-Hubbard model}
\author{Luigi Amico}
\affiliation{MATIS, INFM CNR}
\affiliation{Dipartimento di Metodologie Fisiche e Chimiche (DMFCI),
Universit\`a di Catania, viale A. Doria 6, I-95125 Catania, Italy}
\email{lamico@dmfci.unict.it}
\homepage{http://fisica.ing.unict.it/~lamico}
\author{Andrea Mastellone}
\affiliation{MATIS, INFM CNR}
\affiliation{Dipartimento di Metodologie Fisiche e Chimiche (DMFCI),
Universit\`a di Catania, viale A. Doria 6, I-95125 Catania, Italy}
\email{andrea@femto.dmfci.unict.it}
\homepage{http://fisica.ing.unict.it/~andrea}
\author{Andreas Osterloh}
\affiliation{Institut f\"ur Theoretische Physik, Universit\"at Hannover,
30167 Hannover, Germany}
\email{andreaso@itp.uni-hannover.de}
\homepage{http://www.itp.uni-hannover.de/~andreaso/}
\date{\today}
\begin{abstract}
We study mesoscopic pairing in the one dimensional repulsive
Hubbard model and its interplay with the BCS model in the canonical ensemble.
The key tool is comparing the
Bethe ansatz equations of the two models in the limit of small Coulomb
repulsion. For the ordinary
Hubbard interaction the BCS Bethe equations with infinite pairing
coupling are recovered; a finite pairing
is obtained by considering a further density-dependent
phase-correlation in the hopping amplitude of the Hubbard model.
We find that spin degrees of freedom
in the Hubbard ground state are arranged in a state
of the BCS type, where the Cooper-pairs form an non-condensed liquid on
a ``lattice'' of single particle energies provided by the Hubbard charge
degrees of freedom; the condensation in the BCS ground state corresponds
to Hubbard excitations constituted by a sea of spin singlets.
\end{abstract}
\pacs{74.78.Na,04.20.Jb}
\maketitle
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%% DEFINITIONS %%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\def\bea{\begin{eqnarray}}
\def\eea{\end{eqnarray}}
\def\beq{\begin{equation}}
\def\eeq{\end{equation}}
\def\up{\uparrow}
\def\down{\downarrow}
\def\ket#1{ | #1 \rangle}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%% GREEK LETTERS %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\def\al{\alpha}
\def\be{\beta}
\def\ga{\gamma}
\def\Ga{\Gamma}
\def\de{\delta}
\def\De{\Delta}
\def\ep{\epsilon}
\def\vep{\varepsilon}
\def\la{\lambda}
\def\La{\Lambda}
\def\sig{\sigma}
\def\Sig{\Sigma}
\def\vphi{\varphi}
\def\om{\omega}
\def\Om{\Omega}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%% END OF DEFINITIONS %%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
Intensive analysis has been devoted to trace pairing phenomena in strongly
interacting electronic models\cite{NOZIER,YANG, CAPONE}. Most of the
motivations come from studies in high-$T_c$ superconductivity for which it is
highly desirable that pairing mechanisms are automatically encoded
in the interaction. The Hubbard model constitutes the prototype of the
class of models of interest in the subject\cite{HUBBARD-HIGHTC}.
It describes a lattice gas
of itinerant electrons experiencing an on-site Coulomb interaction.
So far, the evidence of pair-pair correlation is ambiguous
both for repulsive and attractive Hubbard
interaction\cite{MOREO-ZHANG,MARSIGLIO}.
Here we focus on the Hubbard ($1d$) chains, with repulsive
interaction\cite{HUBBARD-1D}.
Not only does a $1d$ analysis
provide a unique theoretical lab to predict the behavior of
compounds in higher dimension, but current technology
actually admits quasi-$1d$ superconductors
to be fabricated; finally, there is
evidence that the stripe ordered domains gives effectively a $1d$ character to the phenomenology of even many superconducting
$3d$-compounds\cite{STRIPED}.
However, how the superconductivity does arise from
stripes still constitutes an open problem.
Assuming that the electronic properties
of the stripe-domains can be captured by repulsive
Hubbard-like models, effective pair-formation can be studied
by means of the pair binding energy $E_{pb}$,
detecting an energy gain whenever an electron-electron attraction
is established into the system\cite{PAIR-BINDING,MARTINS,CHAKRA-KIVELSON}.
In particular, $E_{pb}$ was studied for the ground state (GS)
of the \textit{finite} $1d$ Hubbard model and it turned out that
the repulsive local interaction can be over-screened, leaving an
effective attraction among the electrons\cite{CHAKRA-KIVELSON}.
This phenomenon was called mesoscopic pairing.
%In $d=1$ many electronic models
%are amenable to exact treatments by Bethe Ansatz techniques~\cite{BETHE}.
An important motivation for us is to single out a \textit{pairing
of the BCS type} in the mesoscopic pairing of the Hubbard model.
The attempt to trace BCS pairing within the Hubbard type models is not
new, and it has been
well studied in the physical community, mainly for attractive
interaction. A common feature of such studies\cite{MARSIGLIO}
(with the exception of the variational study in Ref.
\onlinecite{MARSIGLIO-TANAKA}) is that the analysis
was done in the grand-canonical ensemble, involving mean field BCS
(that is exact in thermodynamic limit\cite{BOGOLIUBOV}).
The scenario changes in the canonical ensemble where the BCS gap
vanishes identically, the ``superconducting
phase'' is dominated by quantum fluctuations\cite{MASTELLONE} and the mean
field BCS theory is inadequate\cite{OTHERS}.
On the other hand, the canonical analysis
is a natural tool to study the mesoscopic BCS pairing in the
Hubbard model. This choice in the context of the stripe physics is supported
by various experimental
evidences. Stripe ordered domains in superconducting nano-powders were detected
in Ref. \onlinecite{MOHANTY}; fragmented nano-stripes
($10-20$ nanometers long) were clearly noticed also in bulk high-T
superconductors\cite{NANO-BULK-STRIPES}.
We use the Bethe ansatz solution of the BCS model to compare
the two pictures in the canonical ensemble.
The exact solution of the BCS model was
found by Richardson~\cite{RICHARDSON} in 1963,
but went unnoticed to the condensed matter community.
It was discovered only recently to study
the superconductivity at nanoscale\cite{MASTELLONE,OTHERS}.
The paper is organized in the following manner.
In the next section
%\ref{sec:hubbard-model}
we introduce the Hubbard model
and its Bethe Ansatz due to Lieb and Wu. In section \ref{sec:hubbard-gs}
a \textit{formal} correspondence is established for $U\to 0^+$,
where the spin degrees of freedom (the spin ``quasi-momenta'') in the
Hubbard GS can be seen as
Cooper-pair quasi-energies corresponding to certain excitations of the BCS
model. A finite pairing
coupling can be achieved by considering certain gauge-field
correlated-hopping in the Hubbard model, as shown in the section
\ref{sec:finite-pairing}.
The correspondence of the BCS GS to a ``sea'' of
spin singlets on top of the Hubbard GS, is discussed in section
\ref{sec:bcs-gs}.
The role of the BCS pairing in the mesoscopic pair-binding
is discerned in the section \ref{sec:mesopairing}. Finally,
section \ref{sec:conclusions} is devoted to our conclusions,
and the discussion of some physical implications.
\section{The Hubbard model}
\label{sec:hubbard-model}
We consider a Hubbard chain of length $L$, with $N=N_{\up}+N_{\down}$
particles $M=N_{\down}$ spin down, and periodic boundary conditions.
Formulated in second quantization, the Hamiltonian is
\begin{equation}
\label{eq:Hubbard}
H = - t \, \sum_{j=1 }^L
(c_{j+1,\sigma}^\dagger c_{j,\sigma}+ h.c.)
+ U\, \sum_{ j=1}^Ln_{{j},\uparrow} n_{{ j},\downarrow}\;,
\end{equation}
where $\{c_{j,\sigma}, c_{l,\sigma'}^\dagger \}=\delta_{\sigma,\sigma'} \delta_{j,l}$,
$\{c_{j,\sigma}, c_{l,\sigma'} \}=0$,
and $n_{l,\sigma}:= c_{l,\sigma}^\dagger c_{l,\sigma}$; $\sigma=\{\uparrow,\downarrow\}$ is the electronic spin; $U$ and
$t$ (we set $t=1$) are the Coulomb interaction and the hopping amplitude
respectively.
The many-body wave function
\begin{equation}
\ket {\psi} = \frac{1}{\sqrt{N!}} \sum_{1 \le j_1 \le \dots \le j_N \le L}
\sum_{\mathcal{S}_N} \psi (j_1, \dots, j_N | \pi) c^{\dagger}_{j_1 \sig_{\pi(1)}}
\dots c^{\dagger}_{j_N \sig_{\pi(N)}} \ket{0},
\end{equation}
brings Eq. \eqref{eq:Hubbard} to the first quantized spectral
problem ${\cal H} \psi={\cal E} \psi$, that can be solved by the Bethe
ansatz\cite{FRAHM}
\bea
\psi (j_1, \dots, j_N | \pi) &=& \sum_{Q \in \mathcal{S}_N} A_{\pi}(Q)
\exp \left \{i \sum_{l=1}^N k_{Q(l)} j_l \right \},
\\
A_{\pi}(Q) &=&
\sum_{R \in \mathcal{S}_M} G(R) \prod_{\alpha=1}^M F(\la_{R(\al)},y_\al|Q),
\\
F(\la,y|Q) &=& \frac{2iU}{\la - \sin k_y + iU} \prod_{l=1}^{y-1}
\frac{\la - \sin k_l - iU}{\la - \sin k_l + iU},
\\
G(R) &=& \prod_{1\le \al < \be \le M} \frac{\la_\al - \la_\be -2iU}
{\la_\al - \la_\be}.
\eea
In these equations $\pi$ labels a permutation of the indices in the
configuration space,
whereas $Q$ and $R$ label permutations in the $\{k\}$, $\{\la\}$
values in the symmetric group $\mathcal{S}_N$ and $\mathcal{S}_M$ respectively.
The exact energy and momentum of the model\cite{HUBBARD-1D,FRAHM}
\begin{equation}
E=-2 \sum_{j=1}^N \cos k_j \; \qquad
P=\sum_{j=1}^N k_j %I_j+\sum_{\al=1}^M J_\al
\label{energy}
\end{equation}
are given in terms of charge-momenta $k_j$, solutions of the Lieb-Wu
Bethe Equations (BE) \cite{LIEB-WU}
\begin{eqnarray}
e^{i L k_j} &=& \prod_{\alpha=1}^M
\frac{ \sin k_j-\lambda_\alpha + \frac{iU}{4}}
{ \sin k_j-\lambda_\alpha - \frac{iU}{4} }
\label{eq:BAE1}
\\
\prod_{j=1}^N \frac{ \sin k_j-\lambda_\alpha + \frac{iU}{4}}
{ \sin k_j-\lambda_\alpha - \frac{iU}{4} } &=&
\prod_{\substack{\be=1 \\ \be \neq \al}}^M \frac { \lambda_{\alpha} - \lambda_{\beta}-
\frac{iU}{2} }{ \lambda_{\alpha} - \lambda_{\beta} + \frac{iU}{2} } ,
\label{eq:BAE2}
\end{eqnarray}
where $ j \in \{ 1,2, \dots N\}$ and $\alpha \in \{1,2, \dots M \}$.
The equations \eqref{energy}, \eqref{eq:BAE1}, and \eqref{eq:BAE2}
constitute the exact solution of the
Hubbard model both for repulsive and attractive interactions.
We focus on the limit $U\to 0^+$.
The result of such operation on Eqs.
\eqref{energy}, \eqref{eq:BAE1} and \eqref{eq:BAE2}
depends on the actual configurations of the charge
and spin rapidities $k,\lambda$. For numerics,
systems up to $N=100$ electrons have been considered.
\section{Hubbard ground state for vanishing interaction}
\label{sec:hubbard-gs}
Here we consider the half-filled chain, and $N = 2M$,
$N$ even. For $U>0$ the GS is characterized
by real $k$ and $\lambda $, where any $\lambda_\al$ is trapped between two
$\sin k_\alpha$.
Systems with $N=4n$ electrons are characterized by degenerate
$\sin {k}=0$ (see the central cluster of Fig. \ref{fig:ground} (a)).
At small $U$ the two $\sin k_\al$'s are arranged symmetrically around
$\lambda_\al$ as $|\lambda_j-\sin k_j| \simeq C_j U^{1/2} + D_j U$
where $C_j$ and $D_j$ are numerical coefficients;
from the numerical analysis we find $D_j=1/4$.
% depending on the choice of the two charges.
We further assume the quasi-momenta having the following behavior
$k_j \simeq k_j^{(0)} + U^{1/2} k_j^{(1/2)} + U k_j^{(1)}$;
this assumption will be confirmed by our numerical studies.
The limit $U \rightarrow 0^+$ of Eqs. \eqref{eq:BAE1} and \eqref{eq:BAE2}
reads~\cite{FRAHM,BATCHELOR}
\bea
&& e^{i k_j^{(0)} L} = 1\; ,
\label{eq:group1order0}
\\
%&&[k_j^{(1/2)}]^2 L^2 = \frac{16 D_j^2 + 1}{8 C_j^2},
%\label{eq:group1order1re}
%\\
&&k_j^{(1)} L = \frac{1}{2} \sum_{\alpha} \, '
\frac{1}{\sin k_j^{(0)} -\la_{\al}},
\label{eq:group1order1im} \\
&&\sum_{\be \neq \al} \frac{2}{\lambda_{\beta} - \lambda_{\alpha}}
- \sum_{j=1}^N \frac{1}{\sin k_j - \la_{\al}} = 0.
\label{eq:group2order1}
\eea
and $k_j^{(1/2)} L = {1}/{2 C_j}$.
$\sum'$ in Eq. \eqref{eq:group1order1im} indicates that the term
$\sin k_{\al}$ is omitted.
Notice that $U$ has to be small in order
that both $|U^{1/2} k_j^{(1/2)} L| \ll 1$ and $|U k_j^{(1)} L| \ll 1$
are satisfied and that degenerate $\sin k_j$ may occur in
Eq. \eqref{eq:group2order1} (see Fig. \ref{fig:ground}).
\begin{figure}
\includegraphics[width=7cm]{figure-1a.eps}
\includegraphics[width=7cm]{figure-1b.eps}
\includegraphics[width=7cm]{figure-1c.eps}
\caption{The spin ($\la_{\al}$, black dashed lines)
and charge ($\sin k_j$, red solid lines) rapidities
at $U/t \ll 1$, for a system with 8 (a) and 10 (b) electrons.
A correlated hopping $\Phi_\sigma \neq 0$ removes the degeneration
of the cluster at $k=0$ for $N=4n$ (c).
In the upper cluster, though $\la_\al$ exhibits a slight dependence
on $U$, the functional behavior of the difference $|\la_\al - \sin k_\al|$
remains the same (inset).}
\label{fig:ground}
\end{figure}
The Bethe ansatz wavefunction in the limit of vanishing $U$ is\cite{FRAHM}
\beq
\lim_{U\rightarrow 0} \frac{A_{\pi}(Q)}{(2iU)^M} =
\sum_{R\in \mathcal{S}_M} \prod_{\al=1}^M \frac{1}{\la_{R(\al)}-\sin[k_{Q(y_\al)}]}
\label{eq:wavefun-zerolimit}
\eeq
The relevance of this limit resides in the following observation.
The Eqs. \eqref{eq:group2order1} are known as the Gaudin equations.
They play an important role in the exact solution of the BCS
model at infinite pairing coupling $g$.
The BCS Hamiltonian for $N_p$ time-reversed pairs
in $\Omega$ single particle levels $\varepsilon_j$ with pair degeneracy $d_j$ reads
\cite{SIERRA-RMP}
\bea
\label{BCS-ham}
&&\hspace*{-0.4cm}H_{_{BCS}} =
\sum_m \sum_{j=1}^\Omega \varepsilon_j c^\dagger_{j,m} c_{j,m}-g
\sum_{i,j=1}^\Omega \sum_{m,m'} c^\dagger_{i\,m} c^\dagger_{i\,\bar{m}}
c_{j\,\bar{m'}}c_{j\,m'} \nonumber \\ &&\hspace*{0.5cm} =
\sum_{j=1}^\Omega \varepsilon_j \tau_j \;, \quad
\tau_j = \sum_{i \neq j} \frac{\vec{S}_i \cdot \vec{S}_j}
{\varepsilon_i - \varepsilon_j} + \frac{1}{g} S^z_j,
\eea
where $c^\dagger_{j\,m}$ creates an electron in the state
$\ket{j;S_j,m}$, $m\in\{1\dots d_j\}$;
$|j;S_j,\bar{m}\rangle$ is the time reversed state of $|j;S_j,m\rangle$.
The operators
$
S^-_j := \sum_m c_{j\overline{m}} c_{jm}, \;\; S^+_j:=(S^-_j)^\dagger \;\;
S^z_j := (\sum_{m=-d_j}^{d_j} n_{jm} -d_j)/2$ form a $d_j/2$ dimensional
representation of the $su(2)$ algebra.
The second line of Eq. \eqref{BCS-ham} establishes the relation
between $H_{BCS}$ and the twisted six-vertex models\cite{AMICO-VERTEX}
through the Gaudin model\cite{CAMBIAGGIO}, ultimately
proving the integrability of the BCS model.
Its exact energy and eigenstates\cite{RICHARDSON} are
\begin{eqnarray}
&&\mathcal{E}_{BCS} = \sum_{\alpha = 1}^{N_p} e_\alpha
+ \sum_{i=1}^{N_b} \varepsilon_i \label{BCS-exact-en}\\
&&\frac{1}{g} - \sum_{j=1}^\Omega \frac{d_j}{2 \varepsilon_j -e_\al}+
\sum_{\be\not=\al=1}^{N_p} \frac{2}{e_\be -e_\al}=0 \;,
\label{BAE_BCS}\\
&&|\Psi\rangle =\sum_{R\in \mathcal{S}_{N_p}} \prod_{\al=1}^{N_p} \frac{1}{e_{R(\al)}-
2\varepsilon_{Q(y_\al)}}S^\dagger_{R(\al)}=\prod_{\al=1}^{N_p} \sum_{j=1}^\Omega
\frac{S^\dagger_j}{ e_\al-2 \varepsilon_j } \ket{0} \label{BCS-exact-state}\end{eqnarray}
where %$\al=1\dots N_p$ and
the second sum in Eq.\eqref{BCS-exact-en} runs over the set of $N_b$
blocked levels, i.e. occupied by a single electron
and not available to pair scattering.
%, i.e. the number of Cooper pairs that the level can contain.
%$\Omega_b$ is the number of blocked levels.
Different states in Eq. \eqref{BCS-exact-state} correspond to different
distributions of the $e_\al$ at $g=0$ on the $\varepsilon_j$-configuration;
in the BCS GS all $\varepsilon_j$'s are complex conjugates.
The relation between Eqs. \eqref{BAE_BCS},\eqref{BCS-exact-state}
and Eqs. \eqref{eq:group2order1}, \eqref{eq:wavefun-zerolimit} is evident
identifying $2 \varepsilon_j$ and $e_\al$
(the pairing quasi-energies) with $\sin k_j$ and $\lambda_\al$
respectively,
$\Omega$ and $N_p$ playing the role of $N$ and $M$.
Eqs. \eqref{eq:group2order1} are obtained from Eq. \eqref{BAE_BCS}
in the limit $g\rightarrow \infty$.
Important for our purposes are the excitations
of the BCS model characterized by real $\{ e_\alpha\}$; it was noticed
that in the canonical ensemble these can be obtained as the
states in Eq. \eqref{BCS-exact-state}
corresponding to $\{ e_\alpha\}$, \textit{finite} solutions of the Gaudin
equations\cite{SIERRA-RMP}. Therefore we conclude that the
spin-rapidities (all of them are real and finite) in the Hubbard GS are
arranged along the configuration of the BCS excited state
displayed in the Fig. \ref{fig:bcs-conf}.
\section{Finite pairing}
\label{sec:finite-pairing}
Now we show that a finite value of $g$ can be obtained by adding
%A way we found to obtain a finite value of $g$ is to consider
a further interaction in the
original Hubbard model, in form of correlated-hopping.
The correlated hopping we consider corresponds to
\begin{equation}
t \rightarrow t \exp\Bigl[{\rm i}\sum_{l}^{}\bigl(\alpha_{j,l}(\sigma)N_{l,-\sigma}
+ A_{j,l}(\sigma)N_{l,\sigma}\bigr)\Bigr] \; ,
\label{correlated}
\end{equation}
in Eq. \eqref{eq:Hubbard}.
The Hubbard model with the correlated hopping Eq. \eqref{correlated} are
of the Shastry-Schulz type\cite{SCHULZ-SHASTRY}.
They were solved exactly for $A, \alpha$ obeying certain
restrictions\cite{AMICO-OSTERLOH-ECKERN}.
%\begin{eqnarray}
%\partial_x \alpha(x,y;\sigma) &=& \partial_y \alpha(y,x;\sigma) \\
%\partial_x A(x,y;\sigma) &=& \partial_y A(y,x;\sigma)\ ;\ x\neq y\;,
%\end{eqnarray}
%for consistency ($\partial_x f(x) := f(x+1) -f(x)$ and $(x,y)$ are pair of indices
%in the chain).%~\cite{SHASTRY-SUTEHRLAND}.
%The key to solve the models of the Schulz-Shastry type is to observe that
%the correlated hopping can be gauged away everywhere except at one point of the ring.
It was demonstrated \cite{AMICO-OSTERLOH-ECKERN} that
the effect of the correlated hopping is to twist the boundary conditions.
The boundary phases are
$\Phi_\sigma:=\phi(\sigma)+ \theta_{\uparrow\downarrow}(\sigma)
N_{-\sigma} +\theta_{\uparrow\uparrow}(\sigma)(N_\sigma-1)$,
with $ \phi(\sigma)
=\sum_{j=1}^{L} A_{j,j}(\sigma)$,
$\theta_{\uparrow\downarrow}(\sigma)
= \sum_{j=1}^{L}\alpha_{j,m}(\sigma)$,
$ \theta_{\uparrow\uparrow}(\sigma)
=\sum_{{{j\neq m-1,m}}}^{L} A_{j,m}(\sigma)
+A_{m,m-1}(\sigma)+A_{m-1,m+1}(\sigma)$.
%Eqs. (\ref{INT1}) and (\ref{INT2})
%also ensure that the phases in Eqs. (\ref{P1}) and (\ref{P2}) are
%$m$-independent\cite{NOTEPHASES}.
The BE of the Schulz-Shastry models are obtained from the Lieb-Wu BE,
by multiplying with
the factors $e^{-i\Phi_{\uparrow}}$ and $e^{i(\Phi_{\uparrow}-\Phi_{\downarrow})}$
the r.h.s. of the equations \eqref{eq:BAE1} and \eqref{eq:BAE2} respectively.
To obtain a finite $g$, it is assumed also that
$\Phi_{\sigma} \simeq \Phi_{\sigma}^{(0)} + U \Phi_{\sigma}^{(1)}$.
For $-\pi \le k_j < \pi$ we find that the conditions $\Phi_{\up \down}^{(0)}=
\Phi_{\up}^{(0)} - \Phi_{\down}^{(0)} = 0$ and $|U \Phi_{\up \down}^{(1)}| \ll 1$
must be satisfied, in order to obtain the BE to zero order in $U$.
%Since boundary phase is present in (\ref{eq:BAE1}), notably
The Eqs. \eqref{eq:group1order0} and \eqref{eq:group1order1im} then modify
into
\bea
e^{i k_j^{(0)} L} &=& e^{-i \Phi_{\up}^{(0)}},
\nonumber
\label{eq:newgroup1order0}
\\
k_j^{(1)} L &=& - \Phi_{\up}^{(1)}
+ \frac{1}{2} \sum_{\alpha} \, ' \frac{1}{\sin k_j^{(0)} -\lambda_\alpha},
\nonumber
\label{eq:newgroup1order1im}
\eea
and Eqs.(\ref{eq:group2order1}) into
\begin{equation}
\frac{1}{g} - \sum_{j=1}^N \frac{1}{\sin k_j -\lambda_\al}+
\sum_{\be\not=\al=1}^{M} \frac{2}{\lambda_\be -\lambda_\al}=0 \;,
\label{eq:pairing}
\end{equation}
with
%\begin{equation}
$ g \equiv 1/(2\Phi_{\up \down}^{(1)})$.
%\label{eq:pairing}
%\end{equation}
Thus the BE of the correlated hopping Hubbard model lead,
in the limit $U \rightarrow 0^{+}$, to Richardson BE, Eq. \eqref{BAE_BCS}.
In the same limit the Eq. \eqref{energy} is
$$
E = \frac{2 U \Phi_{\up \down}^{(1)}}{L} \sum_{\al=1}^M \la_{\al}
+ \mathrm{const}.
$$
with the constant being ${U M(N-M+1)}/{L} -2 \sum_{j=1}^N \cos k_j^{(0)}
- ({2 U \Phi_{\up}^{(1)}}/{L}) \sum_{j=1}^N \sin k_j^{(0)}$
That is: for small $U$, the energy of the Hubbard model coincides (up to constants)
with the energy of the BCS model
, see Eq.\eqref{BCS-exact-en}.
%We resume in the following table the corresponding quantities in the %mapping.
%\begin{center}
%\begin{tabular}{|c|c|}
%\hline
%Hubbard & BCS \\
%\hline
%$N$ (Number of particles) & $\Omega$ (Number of levels) \\
%$N_\down=M$ (Number of spin-$\down$ particles) &
%$N_p$ (Number of pairs) \\
%$\sin k_j$ (Quasi-momenta) & $2\varepsilon_j$ (Single particle %energies) \\
%$\la_\al$ (spin rapidities) & $e_\al$ (Quasi-energies) \\
%\hline
%\end{tabular}
%\end{center}
The excitation of the BCS model corresponding to the Hubbard GS
is achieved by filling the levels
$\varepsilon_j$'s with a Cooper pair, leaving one empty
level between two filled ones.
In this way, the empty
levels prevent the pairing parameters to form complex
conjugate pairs. Such a BCS excitation is constituted by an
non-condensed liquid of Cooper pairs (see the discussion in the next
paragraph).
The structure of these excitations is different for $N=4n+2$ and $N=4n$.
In the first case the $\varepsilon_j$ are all non degenerate:
$d_j=1,\; \forall j $; for $N=4n$, instead, the doubly degenerate Fermi
energy correspond to a doubly degenerate $\varepsilon_j$, as one can see in
the Figs. \ref{fig:ground} and \ref{fig:bcs-conf}).
\begin{figure}
\includegraphics[width=7cm]{figure-2a.eps}
\quad
\includegraphics[width=7cm]{figure-2b.eps}
\caption{BCS and Gaudin configurations leading to finite solutions
at large pairing, for $N=4n$ (a) and $N=4n+2$ (b) and $n=2$.
The plot
shows the behavior of the pairing parameters while increasing $g$,
black curves, whereas the single particle energies remain constant,
red lines.
On left sides of the plots, the initial configuration of the
pairing parameters $\la_\al$ (black circles) on the
energy levels $2\varepsilon_j$ (red crosses) is shown;
on the right sides the final configuration is displayed.
The latter are the charge and spin configurations corresponding to
the ground state of the Hubbard model. The larger cross in plot (a)
indicates the double degenerate level $2\varepsilon=0$.}
\label{fig:bcs-conf}
\end{figure}
\section{The BCS ground state}
\label{sec:bcs-gs}
The trait of the condensation of the Cooper pairs in the
BCS GS is the emergence of
complex conjugate $e_\alpha$, solution of Eq. \eqref{BAE_BCS}.
In fact by resorting certain electrostatic analogy
\cite{GAUDIN,RICHARDSON-ELECTRO} it can be proved that
the thermodynamic limit of the Richardson's BE
leads to the gap equation: $2G \int_\Omega \rho(\varepsilon)/
\sqrt{(\varepsilon-a) (\varepsilon-b)}=1$ where $\rho(\varepsilon)$ is the density of single particle levels; the parameters $a=\varepsilon_0+i\Delta$ and
$b=\varepsilon_0-i\Delta$ are the the end points of the arc where the solutions of Eqs.(\ref{BAE_BCS}) are disposed (see
Fig.\ref{fig:pair-parameters-conf}; see also Refs.
\onlinecite{SIERRA-RMP,ELECTRO}).
Thus: the BCS gap (in the thermodynamic limit)
is directly related to the maximum of the imaginary part of the
$e_\alpha$\cite{RICHARDSON-ELECTRO,SIERRA-RMP,ELECTRO}.
\begin{figure}
\includegraphics[width=10cm]{figure-3.eps}
\caption{(a) The real and (b) the imaginary part of the pairing
parameters are shown as functions of the pairing strength $g$
for a system with $N=16$ electrons at half-filling in the BCS
ground state configuration. In the thermodynamic limit the pairing
parameters, being all complex conjugate at sufficiently large pairing,
are arranged along arcs (see Ref. \onlinecite{ELECTRO}),
where the distance between the end-points
is the BCS gap $2\Delta$. Here we have set $g=3.0$ and $N=200$, with
the single particle levels uniformly distributed in the range [-1,1].}
\label{fig:pair-parameters-conf}
\end{figure}
%(i.e. small flux, see eq. \eqref{eq:pairing} ),
For large $g$ all the $e_\alpha$ are complex
conjugated pairs (with a single exception if $N_p$ is odd);
their real part is far below the lowest $\varepsilon_j$
(see e.g. Ref. \onlinecite{SIERRA-RMP}).
To find the Hubbard eigenstate corresponding
to the BCS GS we insert into Eqs. \eqref{eq:BAE1}, \eqref{eq:BAE2}
the $e_\alpha$($\equiv \lambda_\alpha$) configuration
obtained by solving Eq. \eqref{eq:pairing} with certain configuration of
$ \sin k_j^{(0)}$; this fixes the $k_j$, given the $\Phi_\sigma$
we have used for Eq. \eqref{eq:pairing}
with $|U \Phi_{\sigma}^{(1)}| \ll 1$.
We found that all the $k_j$'s
are real and non degenerate for non degenerate $\sin k_j^{(0)}$'s.
(degenerate $\sin k_j^{(0)}$ correspond to degenerate $k_j$).
Numerically we found that the mapping can be reliably done for
$U \lesssim 0.03$. It then turns out that
$\sin k_j=\sin k_j^{(0)} + U \delta_j$;
%for such a state it is always
and $|\sin k-\lambda|\neq 0$ is always satisfied. These conditions
imply that equation (\ref{eq:pairing}) characterizes also the
excited Hubbard state under consideration.
Summarizing, real $k_j$ $\&$ complex conjugate $\la_{\al}$ pairs,
solutions of the Richardson BE, satisfy the Lieb-Wu equations.
We conclude that the BCS GS corresponds to a sea of
spin singlets ($\Lambda$-2 strings) in the Hubbard chain.
Such a state can be seen as a BCS
condensation of spin rapidities.
As an example of such result, the table shows the $k,\lambda$ solution
of both the BCS and the Hubbard GS BE for $U=0.001$, $N=6$, and $M=3$.
\begin{center}
\begin{tabular}{|cc|c|}
\hline
$\sin k_j$ & & $\lambda_\alpha $ \\
\hline
\small
-2.14679391, & \small -1.09959633 & \small -14.8060834 \\
\hline
\small -0.052399128, & \small 0.994798057 & \small -11.722088 - i 11.1576888 \\
\hline
\small 2.04199559, & \small 3.08919348 & \small -11.722088 + i 11.1576888 \\
\hline
\end{tabular}
\normalsize
\end{center}
\section{Mesoscopic pairing}
\label{sec:mesopairing}
Here, we have a look at the pair binding energy
$E_{pb} \doteq 2 E(N+1) -E(N) -E(N+2)$.
For the Hubbard GS,
$E_{pb}$ shows ``super-even effects''\cite{CHAKRA-KIVELSON,PAIR-BINDING}.
Effective pair attraction $E_{pb}>0$ occurs for $N=4n$ ($n$ is integer), and it is
related with a vanishing GS spin gap $\Delta_S\doteq E(N)_{S=1}-E(N)_{S=0}$
for $U \rightarrow 0^+$;
for $N=4n +2$, $E_{pb}<0$, corresponding to $\Delta_S>0$
at small $U$. In Fig.(\ref{fig:bindspin}) we present $E_{pb}$
for the Hubbard and for Shastry-Schulz models. We observe that
the correlated hopping, Eq. \eqref{correlated}, washes out
the super-even effects %, and $\Delta_S>0$ at $U=0$
(see Fig.\ref{fig:bindspin}); for $N=4n+2$ and $U> U_c$ $E_p>0$
(see also Ref. \onlinecite{MARTINS});
at small $N$ we found that $U_c \propto N$. For larger $N$,
$U_c$ decreases (at $N=50$, $U_c\approx 2.5$). The phenomenon will be analyzed elsewhere.
\begin{figure}
\includegraphics[width=8cm]{figure-4a.eps}
\includegraphics[width=8cm]{figure-4b.eps}
\caption{The super-even effects for the pair binding energy $E_pb$ (a) and
the spin gap $\Delta_s$ (b).
Thick and light lines are results for periodic and twisted boundaries
respectively.
$N=10$: black solid line and red dotted-dashed line;
$N=12$: blue dotted line and orange dashed line.
The boundary twist is set to $\Phi^{(0)}_{\up} = \Phi^{(0)}_{\down} = 0.2 \pi$,
$\Phi^{(1)}_{\up} = 0.1 \pi=2\Phi^{(1)}_{\down}$. }
\label{fig:bindspin}
\end{figure}
The formation of $\Delta_S$ can be interpreted
within the Hubbard-BCS correspondence:
the state with $S=1$ is
obtained from $S=0$ by breaking a BCS pair and thereby blocking
single particle levels. In this scenario the super-even effect arises
(see also Fig.\ref{fig:energy-conf}) because breaking a pair
in the state with $N=4n$ is energetically favorable, creating two
blocked degenerate single particle levels, without kinetic energy extra cost.
In contrast, for $N=4n+2$ the blocking of levels occurs at $k>k_F$.
\begin{figure}
\includegraphics[width=4cm]{figure-5a.eps}
\quad
\includegraphics[width=4cm]{figure-5b.eps}
\caption{The energy configuration ($-2\cos k$)
of the charge rapidities in the $S=0,1$ ground states for $N=4n$
(a) and $N=4n+2$ (b) systems.
The different symbols indicate the degeneracy of
the levels: circle, cross and triangle stand for one, two and
three-fold degenerate levels respectively. Note that
the three-fold degenerate levels are more precisely ``quasi-degenerate''
ones: the energy difference cannot be resolved in the plot.
At $N=4n$, passing from $S=0$ to $S=1$ state two degenerate single
particle levels are created and no additional energy is required (as
shown by the arrows). Instead, at $N=4n+2$, at half filling the
levels are reshuffled, a pair is broken at Fermi level and a $k=0$
level is occupied (upper arrow).}
\label{fig:energy-conf}
\end{figure}
Away from half filling this is sufficient to cause an increase of the total energy;
at half filling the levels are blocked
in such a way that all the levels in the band are reshuffled,
since breaking a pair at the Fermi energy causes the occupancy of the state at $k=0$.
We study the phenomenon by considering the pair
correlation function $\Psi_j=|u_j|$, where
$u_j\doteq \langle c^\dagger_{j\,m} c^\dagger_{j\,\bar{m}} c_{j\,\bar{m}}c_{j\,m}\rangle $
detecting the fluctuational superconductivity in the canonical ensemble
(in the thermodynamic limit $\Psi_j$ becomes the BCS gap)\cite{MASTELLONE}.
Here we exploit the exact results achieved in
Ref. \onlinecite{BCS-CORR} where $\Psi_j=|u_j|$ is calculated
from a generating function\cite{SKLYANIN-CORR}.
Exclusively for
systems with pair attraction ($N=4n$ with $\Phi_\sigma\equiv 0$)
a dip at $k=0$ is observed (Fig.\ref{fig:paircorr}); this reflects the
corresponding small energy gap required to occupy the $k=0$ mode in
the $S=0\rightarrow S=1$ process.
\begin{figure}
\includegraphics[width=10cm]{figure-6.eps}
\caption{The pairing correlator $\Psi_j=|u_j|$
is presented (in the thermodynamic limit $\Psi_j$ would become the BCS gap).
We notice the anomaly for $N=4n$.}
\label{fig:paircorr}
\end{figure}
\section{Conclusions}
\label{sec:conclusions}
We proved that the exact solution of the BCS model
can be obtained from the Bethe ansatz solution of the Hubbard model at
vanishing $U$. At a formal level we comment that the amplitudes $A_\pi(Q)$ diagonalizing the 'spin sector' of the Hubbard model are, in fact, eigenvectors of the transfer matrix of certain inhomogeneous six vertex models\cite{FRAHM}.
The relation between the Hubbard chain and the BCS model arises
because the BCS model is itself a quasi-classical descendent of
the inhomogeneous six vertex model\cite{AMICO-VERTEX}.
The Coulomb repulsion $U/t$ plays the role of the quasi-classical parameter;
the charge degrees of freedom $\{\sin k \}$ of the Hubbard model play the role
of the inhomogeneities for the vertex models. In the
BCS picture, they provide
the ``lattice'' of single particle energies $\varepsilon_j$ where the spin
degrees of freedom, the spin ``quasi-momenta'' $\lambda_\alpha$,
can condense.
The Hubbard ground state corresponds to a certain
excited BCS state (see Fig.\ref{fig:bcs-conf}).
We noticed that also the Hubbard low lying spin excitations corresponds
to states of the BCS type.
In particular we have demonstrated that the BCS ground state of the spin rapidities
reflects the formation of a sea of spin singlets in the Hubbard chain.
In this scenario
the super-even effective attraction in the Hubbard and in the Shastry-Schulz
ground states can be understood in terms of pair-breaking excitations
in the BCS model. We have demonstrated that BCS-correlators serve to study
such parity effects.
The results obtained in this paper might be relevant for the analysis of
the stripe order in the high-$T_c$ compounds.
\begin{acknowledgments}
A. Di Lorenzo, F. Dolcini, G. Falci, R. Fazio, H. Frahm, A. Fubini,
M. Roncaglia, and G. Sierra are
acknowledged for discussions and support.
\end{acknowledgments}
\begin{thebibliography}{99}
\bibitem{NOZIER} A.J. Leggett, J. de Physique, C7, \textbf{ 41}, 19 (1980);
P. Nozi\'eres and S. Schmitt-Rink, J. Low Temp. Phys. \textbf{ 59}, 195 (1985).
\bibitem{YANG} C.N. Yang, Phys. Rev. Lett. \textbf{ 63}, 2144 (1989).
\bibitem{CAPONE} M. Capone, \textit{et al.}, Science \textbf{ 296}, 2364 (2002).
\bibitem{HUBBARD-HIGHTC} E.W. Carlson, \textit{et al.},
\textit{``The Physics of Conventional and Unconventional Superconductors.''}
ed. by K. H. Bennemann and J. B. Ketterson (Springer-Verlag, Berlin, 2000)
\bibitem{MOREO-ZHANG}A. Moreo, Phys. Rev. B \textbf{ 45}, 5059 (1992).
%S.-C. Zhang, J. Carlson, and J.E. Gubernatis, Phys. Rev. B \textbf{ 55},
7464 (1997).
%\bibitem{QUICK} R.M. Quick, C. Esebbag and M. de Llano,
%Phys. Rev. B \textbf{ 47}, 11512 (1993).
\bibitem{MARSIGLIO} F. Marsiglio, Phys. Rev. B \textbf{ 55}, 575 (1997);
W. Fettes, I. Morgenstern, and T. Husslein, Int. J. Mod. Phys. C
\textbf{ 8}, 1037 (1997).
%\bibitem{FETTES} W. Fettes, I. Morgenstern, and T. Husslein,
%Int. J. Mod. Phys. C8, 1037 (1997).
\bibitem{HUBBARD-1D}N. Andrei, \textit{``Integrable models in condensed matter
physics''}, in ``Low dimensional quantum field theories for condensed
matter physicist'', ed. S. Lundqvist, G. Morandi, and Yu Lu
(World Scientific, Singapore, 1995).
\bibitem{STRIPED} X. J. Zhou \textit{et al.}, Science \textbf{ 286}, 268 (1999).
\bibitem{PAIR-BINDING}V.J. Emery, S.A. Kivelson, and H.Q. Lin,
Phys. Rev. Lett. \textbf{ 64}, 475 (1990).
\bibitem{MARTINS} R.M. Fye, M.J. Martins, and R.T. Scalettar,
Phys. Rev. B \textbf{ 42}, 6809 (1990).
\bibitem{CHAKRA-KIVELSON} S. Chakravarty and S. A. Kivelson,
Phys. Rev. B \textbf{ 64}, 064511 (2001).
\bibitem{MARSIGLIO-TANAKA} K. Tanaka and F. Marsiglio,
Phys. Rev. B \textbf{ 60}, 3508 (1999);
see also N. Salwen, S.A. Sheets, and S.R. Cotanch, Phys. Rev. B \textbf{70},
064511 (2004).
\bibitem{BOGOLIUBOV} N.N. Bogoliubov, Nuovo Cimento \textbf{7}, 794 (1958).
\bibitem{MASTELLONE}A. Mastellone, G. Falci, and R. Fazio,
Phys. Rev. Lett. \textbf{ 80}, 4542 (1998).
\bibitem{OTHERS}
% K.A. Matveev and A.I. Larkin, Phys. Rev. Lett. \textbf{ 78}, 3749 (1997);
J. von Delft and D.C. Ralph, Phys. Rep. \textbf{ 345}, 61 (2001).
\bibitem{MOHANTY} P. Mohanty \textit{et al.}, Physica C \textbf{ 408}, 666 (2004).
\bibitem{NANO-BULK-STRIPES} C.C. Tsuei and T. Doderer, The European
Physical Journal B \textbf{ 10}, 257 (1999); C.J. Chen and C.C. Tsuei,
Solid State Commun. \textbf{ 71}, 33 (1989); H.L. Edwards \textit{et al.},
Phys. Rev. Lett. \textbf{ 73}, 1154 (1994); H.L. Edwards, \textit{ibidem}
\textbf{75}, 1387 (1995); S. Mori, C.H. Chen, and S.-W. Cheong,
Phys. Rev. Lett \textbf{ 81}, 3972 (1998).
\bibitem{RICHARDSON}R.W. Richardson and N. Sherman, Nucl. Phys.
\textbf{ 52}, 221 (1964); ibid., \textbf{ 52}, 253 (1964).
\bibitem{FRAHM} F. H. L. Essler, \textit{et al.}
%H. Frahm, F. G\"ohmann, A. Kl\"umper, and V.E. Korepin,
\textit{The One-Dimensional Hubbard Model}, (Cambridge University Press, 2005).
\bibitem{LIEB-WU} E.H. Lieb and F.Y. Wu, Phys. Rev. Lett. \textbf{ 20} 1445 (1968).
\bibitem{BATCHELOR} For the same limit in the Bethe equations of
the $\delta$-interacting bose-gas see M.T. Batchelor, X.W. Guan,
and J.B. McGuire,
J. Phys. A \textbf{ 37} (2004) L497; M.T. Batchelor, \textit{et al.},
J. Phys. Soc. Jpn. \textbf{ 74} Suppl. 53-56 (2005).
\bibitem{SIERRA-RMP} J. Dukelsky, S. Pittel, G. Sierra,
Rev. Mod. Phys. \textbf{ 76}, 643 (2004).
\bibitem{AMICO-VERTEX} L. Amico, G. Falci, and R. Fazio, J. Phys. A
\textbf{ 34}, 6425 (2001); A. Di Lorenzo \textit{et al}.,
Nucl. Phys. B, \textbf{644}, 409 (2002).
\bibitem{CAMBIAGGIO} M.C. Cambiaggio, A.M.F. Rivas, and M. Saraceno,
Nucl. Phys. A \textbf{ 624}, 157 (1997).
%\bibitem{SIERRA-EXCITATION}J. M. Roman, G. Sierra, and J. Dukelsky,
%Phys. Rev. B \textbf{ 67}, 064510 (2003)
\bibitem{SCHULZ-SHASTRY} H.J. Schulz and B.S. Shastry,
Phys. Rev. Lett. \textbf{ 80}, 1924 (1998).
\bibitem{AMICO-OSTERLOH-ECKERN} A. Osterloh, L. Amico and U. Eckern,
Nucl. Phys. B \textbf{588}, 531 (2000).
\bibitem{GAUDIN} M. Gaudin ``Travaux de M. Gaudin: Mod\`eles
Exactement r\'esolus'', Les Editions de Physique, France, 1995.
\bibitem{RICHARDSON-ELECTRO} R.W. Richardson, J. Math. Phys. \textbf{ 18},
1802 (1977).
\bibitem{ELECTRO} L. Amico \textit{et al}., Annals of Physics \textbf{ 299},
228 (2002).
\bibitem{BCS-CORR} L. Amico and A. Osterloh, Phys. Rev. Lett. \textbf{ 88},
127003 (2002).
\bibitem{SKLYANIN-CORR} E. Sklyanin, Lett. Math. Phys. \textbf{47}, 275 (1999).
%\bibitem{BETHE} V. E. Korepin and F.H.L. Essler,
%\textit{Exactly solvable models of strongly
%correlated electrons}, (World Scientific, Singapore 1994).
%\bibitem{SKLYANIN-GAUDIN}E.K. Sklyanin, J. Sov. Math. \textbf{ 47}, 2473 (1989).
%\bibitem{SHASTRY-SUTHERLAND} B.S. Shastry and B. Sutherland,
%Phys. Rev. Lett. \textbf{65}, 243 (1990).
%\bibitem{GAUDIN} M. Gaudin, J. Phys. \textbf{ 37}, 1087 (1976).
%\bibitem{KULIK}I. O. Kulik,
%Sov. Superconductivity: Phys. Chem. Tech. \textbf{ 2}, 201 (1989);
%H. Boyaci and I. O. Kulik, Low Temp. Phys. \textbf{ 24}, 239 (1998).
%\bibitem{HIRSCH} J.E. Hirsh, Phys. Lett. A, \textbf{ 134}, 451 (1989);
%J.E. Hirsh and F. Marsiglio, Phys. Rev B, \textbf{ 39}, 11515 (1989);
%J.E. Hirsh, Phys. Rev B, \textbf{ 48}, 3327 (1993);
%Phys. Rev B, \textbf{ 48}, 3340 (1993); Phys. Rev B, \textbf{ 48}, 9815 (1993).
%\bibitem{KETTERLE} A.E. Leanhardt, \textit{et al.} Phys. Rev. Lett.
%\textbf{ 29}, 190403 (2002).
\end{thebibliography}
\end{document}