Realization of odd-frequency p-wave spin-singlet superconductivity coexisting with antiferromagnetic order near quantum critical point
Extracted source text ยท arXiv cond-mat/0302532
\documentclass[preprint]{jpsj2}
%%%\documentclass[twocolumn,seceq]{jpsj2}
%%%\usepackage[mtbold,mtpluscal,T1]{mathtime}
\def\runtitle{Instructions for the Preparation of Manuscript}
\def\runauthor{Yuki \textsc{Fuseya},
Hiroshi \textsc{Kohno} and Kazumasa \textsc{Miyake}}
\title{%
Realization of Odd-Frequency $p$-Wave Spin-Singlet
Superconductivity Coexisting with Antiferromagnetic Order near
Quantum Critical Point
}
\author{%
Yuki \textsc{Fuseya},
Hiroshi \textsc{Kohno} and
Kazumasa \textsc{Miyake}
}
\inst{%
Division of Materials Physics,
Department of Physical Science, Graduate School of Engineering Science, \\
Osaka University, Toyonaka, Osaka 560-8531, Japan
}
\recdate{\today}
\abst{It is shown that the $p$-wave singlet superconductivity, whose gap
function is odd both in momentum and in Matsubara frequency, is realized
prevailing over the $d$-wave singlet state very close to
the antiferromagnetic quantum critical point (QCP) both in the paramagnetic
and in antiferromagnetic sides. Off the QCP in the paramagnetic phase, however,
the $d$-wave singlet superconductivity with line-nodes is realized
as \textit{conventional} anisotropic superconductivity.
This $p$-wave singlet superconductivity is an essentially gapless state,
i.e., there is no gap in the quasiparticle spectrum everywhere
on the Fermi surface due to its
odd-frequency gap. These features can give a qualitative
but nice understanding of the anomalous behaviors of
NQR relaxation rate on CeCu$_2$Si$_2$ or CeRhIn$_5$
where the antiferromagnetism and superconductivity
coexist on a microscopic level.
}
\kword{$p$-wave singlet superconductivity, gapless superconductivity,
coexistence of antiferromagnetism and superconductivity,
CeCu$_2$Si$_2$, CeRhIn$_5$
}
\newcommand{\Tc}{T_{\textrm{c}}}
\newcommand{\siml}{\stackrel{<}{_\sim}}
\newcommand{\simg}{\stackrel{>}{_\sim}}
\begin{document}
\sloppy
\maketitle
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
%==============================================================================
The discovery of the superconductivity in CeCu$_2$Si$_2$ \cite{Steglich}
two decades ago
triggered off a breakthrough in the field of superconductivity.
%
Nowadays, we seem to have reached the consensus that
the anisotropic spin singlet pairing, e.g. ``$d$-wave
singlet" pairing, is realized in the superconducting (SC) states
close to the antiferromagnetic (AF) states due to the AF spin fluctuations.
\cite{MSV,Scalapino,Monthoux,Ueda}
%
Recent developments on CeCu$_2$Si$_2$, however, suggest that
it still disguises much various physics.
\cite{Onishi,Ishida,YKawasaki1,YKawasaki2,Koda,SO5,Kitaoka1,Yuan}
%
Polycrystalline sample of Ce$_{0.99}$Cu$_{2.02}$Si$_2$ shows a peak
in the nuclear spin lattice relaxation rate $1/T_1 $ at
SC transition temperature $\Tc \sim 0.65$K
and $1/T_1 \propto T$ behavior at $T \leq \Tc$,
which indicates essentially gapless superconductivity
to be realized at $T < \Tc$ at ambient pressure.
%
At pressures $P \simg 0.1$GPa, however, the $1/T_1 \propto
T^3 $ behavior, indicating line-node gap, is observed.
\cite{Ishida,YKawasaki1,SO5}
%
Since the Ge substitution for Si is considered to expand the lattice
constant, giving negative pressure, 1\% Ge substituted compound
CeCu$_2$(Si$_{0.99}$Ge$_{0.01}$)$_2$ exhibits
AF ordering at $T< T_{\textrm N}=0.75$K, followed by the onset
of superconductivity at $T=\Tc =0.5$K at ambient pressure.
%
The observation of single component of NQR signal, showing an
appearance of the internal field throughout the sample,
excludes the possibility of phase segregation between the SC
and the AF phases; namely, SC and AF coexist on a microscopic level.
%
In this SC states, $1/T_1$ does not show any significant reduction
below $\Tc$, but shows essentially the same behavior
as the normal Fermi liquid state.
%
At $P=0.85$GPa, on the other hand, it does not exhibit the AF ordering
and the line-node SC gap evidenced from a clear behavior of
$1/T_1 \propto T^3$
\cite{YKawasaki2}.
%
Interestingly, the recently discovered pressure induced superconductor,
CeRhIn$_5$\cite{Heeger} of Ce-based heavy fermion,
shows features quite similar to those of CeCu$_2$Si$_2$,
while its crystal structure is rather two-dimensional
compared to CeCu$_2$Si$_2$.
%
At ambient pressure, CeRhIn$_5$ shows AF ordering at $T<T_{\textrm N}=3.8$
K and $T_{\textrm N} $ exhibits a moderate variation for
$P\siml 1.75$GPa, while
the internal field $H_{\textrm{int}}$ is linearly reduced in the region
$0<P<1.23$GPa, suggesting that $H_{\textrm{int}}$ vanishes
at $P\simeq 1.6$GPa.
%
At $P \sim 1.6$GPa, the SC state begins to appear,
even though the AF ordering still exists.
\cite{Heeger,Mito,SKawasaki,Kitaoka1}
%
The detailed NQR measurement shows that
at $P=1.75 $GPa, the AF order and the SC order
coexist on a microscopic level and $1/T_{1}$ in SC state exhibits
essentially the same $T$-dependence as the normal state
as is seen in CeCu$_2$Si$_2$.\cite{Mito,SKawasaki,Kitaoka1}
%
Under the pressures $P>2.1$GPa, where the AF state disappears,
the $1/T_1 \propto T^3$ behavior
is observed as a usual singlet superconductivity with line-nodes.
\cite{Mito}
%
The specific heat of CeRhIn$_5$\cite{Fisher}
exhibits a feature consistent with the NQR measurement.
%
At $P\sim 1.6$GPa, the specific heat shows a broad peak which is
expected to be associated with the AF ordering at $T\sim 3$K,
and shows a shoulder at $T\sim 2$K, which indicates the appearance of the
gapless superconductivity.
%
This shoulder grows to a mean-field like peak at $P\geq 1.9$GPa,
which would
be associated with the anisotropic singlet pairing.
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{phase.eps}
\caption{Schematic phase diagram near the phase boundary between
antiferromagnetic and superconducting states based on the NQR
experiments on CeCu$_2$Si$_2$ and CeRhIn$_5$.
}\label{phase}\end{center}\end{figure}
Putting all these together, we can draw a schematic phase diagram
near the phase boundary between AF and SC states
as shown in Fig.\ref{phase}.
%
The $P$-$T$ phase diagram can be separated into three parts:
AF phase, SC phase and a coexisting phase of AF and SC ordering
(AF+SC phase).
%
This coexistence is quite different from that of
UPd$_2$Al$_3$,
in which plural $f$ electrons,
in $(5f)^3$-configuration of U$^{3+}$ ion,
shows a dual character of correlated electrons.
%
Of three $f$-electrons per U$^{+3}$ ion,
two electrons form essentially localized $5f^2$-electron state,
as in the U$^{4+}$ state with singlet crystalline electric field ground
state,
while the remaining one electron forms the heavy Fermi liquid
due to a larger
hybridization with conduction electrons.\cite{Sato}
%
In Ce-based heavy fermion compounds, on the other hand,
there is only one $4f$-electron
per Ce$^{3+}$ ion, and thus the same $f$-electron exhibits
simultaneously itinerant and localized
dual nature.\cite{Kuramoto}
%
The qualitative difference of the manner of the coexistence of AF and SC
can easily be seen by the property of $1/T_1$ on UPd$_2$Al$_3$\cite{Tou}
and CeCu$_2$Si$_2$.\cite{Ishida,YKawasaki1,YKawasaki2}
%
A crucial aspect of the AF+SC phase is that
the gapless superconductivity is realized,
although the singlet pairing with line-node due to the
antiferromagnetic fluctuation is expected to emerge.
%
Although one can soon suspect that the origin of this gapless
superconductivity is due to the impurity scattering,
all the samples above are expected to be very clean.
%
In fact,
the clear $1/T_1 \propto T^3$ behavior is observed on the same sample
under the pressures away from the boundary between AF+SC and SC phases.
%
Even if the sample is very clean, there is still a possibility of the
enhanced impurity scattering associated with quantum critical phenomena.
\cite{Kotliar,Narikiyo,Maebashi}
%
In such a case, however, $\Tc$ should be reduced considerably,
while Ce$_{0.99}$Cu$_{2.02}$Si$_2$
at ambient pressure, exhibiting $1/T_1 \propto T$,
does not show a clear decrease of $\Tc$.\cite{Ishida,YKawasaki1,SO5}
%
Therefore, the AF+SC phase in these Ce-based compounds seem to show
the possibility of novel mechanism of ``unconventional" superconductivity.
%
In this paper, we present a possible scenario to unravel these
mysterious features observed in CeCu$_2$Si$_2$ and CeRhIn$_5$
near the boundary between AF+SC and SC phases.
%
An essential idea here is that
a gapless $p$-wave singlet superconductivity with
the so-called odd-frequency gap,
which is odd both in momentum and in frequency,
should be realized very near the quantum critical point (QCP) and/or in
the AF+SC phase.
%
It depends on the specific property of the systems wheather the boundary
between AF+SC phase and SC phase corresponds to QCP or not.
%
However, the present theory can apply to both cases.
%
The Pauli principle requires that the spin singlet gap function
must be even under simultaneous
parity- and time-inversion operations.
%
Therefore, there are two types of singlet gap function;
the first one is even both in frequency and in momentum,
the other is odd both in frequency and in momentum.
%
The latter type of gap was proposed by
Balatsky and Abrahams (BA) in the context of the research
of high-$\Tc$ cuprates about a decade ago\cite{Balatsky},
after the proposal by Berezinskii for the model of superfluid
$^3$He.\cite{Berezinskii}
%
BA argued that there is no gap in the quasiparticle
spectrum with odd-frequency gap and these superconductors
would still exhibit the Meissner effect,
just as in the case of odd-energy gap
proposed at very early stage of research
as a model of CeCu$_2$Si$_2$ which was claimed to exhibit
gapless nature of $1/T_1 $.\cite{Matsuura,Kitaoka2}
%
The model phonon interaction discussed by BA was shown to be
an impractical one,
and it was speculated without specifying any explicit models
that electron-electron interaction might mediate
such odd frequency pairing in general.\cite{Abrahams}
%
Here we show that a realistic model interction mediated by
the critical spin fluctuatioin can give rise to the odd-frequency
gap superconductivity if the Fermi surface of quasi particles
satisfies some mild condition.
%
In the next section, we describe the model and formalism of
solving a linearized gap equation to obtain
the SC transition temperature $\Tc$.
%
In \S 3, we present the numerical results and discuss
the condition that $\Tc$ of $p$-wave singlet superconductivity
prevails over that of $d$-wave singlet superconductivity.
%
In \S 4, we show that the obtained $p$-wave singlet SC state
is really a gapless state also in real frequency space,
i.e., $\Delta (\omega =0)=0$.
%
In \S 5, we apply the same method to the region where the AF
and SC order coexist and show the $p$-wave singlet superconductivity
is also realized there.
%
Then, we discuss the correspondence between
results of our theory and experimental results.
%
Section 6 is devoted to sammary.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Model and Formulation}
%==============================================================================
First we introduce the effective interaction taking into account
the characteristic features of Ce-based compounds CeCu$_2$Si$_2$
and CeRhIn$_5$ near the boundary between the AF and SC phase.
%
For this purpose, it may be useful to recall the fact that
the SC transition temperature of CeCu$_2$Si$_2$ (CeRhIn$_5$) exhibits
maximum at away from the phase boundary $P^* \sim 2.5$GPa
\cite{Thomas,Holmes} (2.5GPa\cite{}).
%
It was shown theoretically, on the standard approximation for an
extended Anderson lattice model, that
the maximum of $\Tc$ is raised by the critical
valence fluctuations\cite{Onishi}.
%
The recent experiment on CeCu$_2$(Si$_{0.9}$Ge$_{0.1}$)$_2$
has revealed that the SC phase is
separated into the phase near the phase boundary of AF phase
and that around $P\sim P^*$\cite{Yuan}.
%
So, it is natural to consider that the latter is due to the critical
valence fluctuations and the former is due to
the critical spin fluctuations.
%
This would be also valid for the CeRhIn$_5$ which exhibits
the analogous features with the CeCu$_2$Si$_2$.\cite{private}
%
Upon this consideration, in order to discuss the mechanism of
superconductivity near the phase boundary,
we introduce the effective interaction
$V(\textbf{q}, \textrm{i} \omega_m)$
as the following phenomenological form mediated by spin fluctuations:
\begin{eqnarray}
V(\textbf{q}, \textrm{i} \omega_m)=g^2\chi(\textbf{q}, \omega_m)
\equiv \frac{g^2 N_F}{\eta + A\hat{\textbf{q}}^2 +C|\omega_m |},
\label{pairint}
\end{eqnarray}
where $g$ is the coupling constant, $N_F$ the densiyy of states
at the Fermi level, and $\hat{\textbf{q}}^2\equiv 4+2(\cos q_x +\cos q_y)$
in two dimensions.
%
This type of pairing interaction was adopted by Monthoux and Lonzarich to
discuss the strong coupling effect on the superconducticity induced
by the critical AF fluctuations.\cite{Monthoux}
%
One of the motivations of adopting $\hat{\textbf{q}}^2$ as above is
that the ordering vector $\textbf{Q}$ of AF phase in CeRhIn$_5$
was identified as $\textbf{Q}=(1/2, 1/2, 0.297)$ by the
neutron diffraction measurements,\cite{Bao}
while that in CeCu$_2$(Si$_{0.99}$Ge$_{0.01}$)$_2$ has not been
identified yet.
%
Another one is that calculations become much simpler in two dimensions
than three dimensions while qualitative physical picture is remained
unchanged.
%
The linearized gap equation is given in a weak-coupling approximation
as follows:
\begin{eqnarray}
\Delta (\textbf{k}, \textrm{i}\varepsilon_n)
=-T\sum_{\textbf{k}',\varepsilon_n '}
\frac{V (\textbf{k}-\textbf{k}',
\textrm{i}\varepsilon_n -\textrm{i}\varepsilon_n ')}
{\xi_{\textbf{k}'}^2 + |\varepsilon_n '|^2}
\Delta (\textbf{k}', \textrm{i}\varepsilon_n '),
\end{eqnarray}
where $\xi _{\textbf{k}}$ is the dispersion of the quasiparticle,
and the pairing interaction is given by (\ref{pairint}).
%
The pairing interaction can be decomposed as
\begin{eqnarray}
V (\textbf{k}-\textbf{k}', \textrm{i}\omega_m)=
\sum_l V_l (\textrm{i}\omega_m )\phi_l^* (\textbf{k})\phi_l (\textbf{k}'),
\label{irr}
\end{eqnarray}
where $\phi_l (\textbf{k})$ is an irreducible representation
of crystal group.
%
%
The coefficient $V_l (\textrm{i}\omega_m)$ in (\ref{irr})
is given as
\begin{eqnarray}
V_l (\textrm{i}\omega_m)=\sum_{\textbf{k}, \textbf{k}'}
\phi_l (\textbf{k}) V (\textbf{k}-\textbf{k}', \textrm{i}\omega_m)
\phi_l ^*(\textbf{k}') .
\end{eqnarray}
%
In order to make our argument as simple and transparent as
possible, we concentrate on the frequency dependence of the
gap function and assuming the two-dimensional circular Fermi surface
shown in Fig. \ref{FS}.
%
Then, the $p$- and $d$-wave components are retained:
\begin{eqnarray}
V_l (\textrm{i} \omega_m ) &=& \int \frac{d\textbf{k}}{v_F }
\frac{d\textbf{k}'}{v_F }
\phi_l (\textbf{k} ) V(\textbf{k} -\textbf{k}', \textrm{i}\omega_m )
\phi_l^* (\textbf{k}), \label{intVl}\\
\phi_p (\textbf{k}) &\equiv& \Phi_p^{-1}N_F^{-1}
\delta( \xi _{\textbf{k}}-\mu ) \sin k_{Fx}
\,\,(\equiv z_p^{-1}N_F^{-1}\delta( \xi _{\textbf{k}}-\mu )\sin k_{Fy} ),\\
\phi_d (\textbf{k}) &\equiv& \Phi_d^{-1}N_F^{-1}
\delta( \xi _{\textbf{k}}-\mu )(\cos k_{Fx} - \cos k_{Fy}),
\end{eqnarray}
where the wave vector on the Fermi surface are expressed as
$\textbf{k}_F = (k_{Fx}, k_{Fy})$,
and $\Phi_l$ is the normalization factor.
The linearized gap equation for each partial-wave component
can be written in the form:
\begin{eqnarray}
\lambda (T)\Delta_l (\textrm{i}\varepsilon_n)
=-T\sum_{\textbf{k}',\varepsilon_n '}
\frac{V_l ( \textrm{i}\varepsilon_n -\textrm{i}\varepsilon_n ')}
{\xi_{\textbf{k}'}^2 + |\varepsilon_n '|^2}
\Delta_l (\textrm{i}\varepsilon_n ').
\label{gapeq}
\end{eqnarray}
%
Assuming $\xi_{\textbf{k}}=v_F (|\textbf{k}|-k_F )$,
one can carry out the integration
of eq. (\ref{gapeq}) with respect to $\textbf{k}'$
as follows:
\begin{eqnarray}
\sum_{\textbf{k}}
\frac{1}{\xi_{\textbf{k}}^2 + |\varepsilon_n |^2}
&=&\frac{1}{(2\pi )^2}\int \textrm{d}\textbf{k}
\frac{1}{v_F^2 (|\textbf{k}|-k_F)^2 +|\varepsilon_n |^2} \nonumber \\
&\simeq &\frac{1}{2\pi }\int_{k_F -k_0}^{k_F + k_0} \textrm{d}k_F
\frac{k}{v_F^2 (k-k_F)^2 +|\varepsilon_n |^2} \nonumber \\
&=&\frac{k_F}{\pi v_F |\varepsilon_n |}
\tan ^{-1} \frac{v_F k_0}{|\varepsilon_n |}.
\end{eqnarray}
As mentioned above,
the gap function must satisfy the following symmetry
for the spin-singlet pairing
\begin{eqnarray}
\Delta_d (\textbf{k}, \textrm{i}\varepsilon_n)
&=&\Delta_d (-\textbf{k}, \textrm{i}\varepsilon_n)
=\Delta_d (\textbf{k}, -\textrm{i}\varepsilon_n)
=\Delta_d (-\textbf{k}, -\textrm{i}\varepsilon_n), \label{even}\\
\Delta_p (\textbf{k}, \textrm{i}\varepsilon_n)
&=&-\Delta_p (-\textbf{k}, \textrm{i}\varepsilon_n)
=-\Delta_p (\textbf{k}, -\textrm{i}\varepsilon_n)
=\Delta_p (-\textbf{k}, -\textrm{i}\varepsilon_n). \label{odd}
\end{eqnarray}
%
Here we measure energies in unit of the Fermi energy $\varepsilon_F $.
%
$g^2 N_F $ is set to be equal to $\varepsilon_F$ throughout the paper.
%
The Fermi surface (FS) is assumed to be two-dimensional circular
with $k_F = 0.75 \pi$ as is shown in Fig. \ref{FS}.
%
This simplified model may give clear understanding of a condition
for the realization
of the odd-frequency gap, although it does not corresponds to
the practical Fermi surface of the heavy fermion compounds.
%
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.5\linewidth]{FS.eps}
\caption{Relation between the Fermi surface ($k_F = 0.75 \pi$)
and the ordering vector $\textbf{Q}$.
}\label{FS}\end{center}\end{figure}
%
The transition temperature $\Tc$ is determined by condition
$\lambda (\Tc )=1$.
%
We have solved the eigenvalue problem (\ref{gapeq}) numerically
by retainning 512 Matsubara frequency $\varepsilon _n$
up to $|\varepsilon _n |\le 131583\pi T$.
%
The temperature dependences of the eigen value $\lambda (T)$'s
are shown in Fig. \ref{lambda}.
%
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{lambda.eps}
\caption{Temperature dependence of the eigen value $\lambda $
for $p$-wave singlet ($p$-S) and $d$-wave singlet ($d$-S).
}\label{lambda}\end{center}\end{figure}
%
For the $d$-wave pairing, $\lambda (T)$ increases monotonously
(logarithmically) as the temperature decreases as usual.
%
For the $p$-wave odd-frequency pairing, on the other hand, $\lambda (T)$
exhibits a broad peak as shown in Fig. \ref{lambda}.
%
Since this feature can be seen even in the case of the much larger
cut-off frequency, it is not due to
the cut-off frequency, but due to an intrinsic nature of the odd-pairing
in frequency.
%
Consequently, it is possible that there exist two temperatures $\Tc ^{\pm}$
satisfing $\lambda (\Tc ^{\pm})=1$,
so that the $p$-wave pairing is realized only in the temperature
region $\Tc ^- < T<\Tc ^+$ (see the case
$\eta = 0.05$ in Fig. \ref{lambda}).
%
But in the case $\eta < 0.04$, $\lambda$ does not fall below unity
even for $T\to 0$, so that the $p$-wave pairing is realized
in $0\le T<\Tc ^+$.
%
We recall that BA reached the anologous results
for the odd-frequency pairing
that the normal phase reappears below the lower-$\Tc$.
%
In the present model, we cannot show the existence of $\Tc^{\pm}$,
since
the transition temperature for $d$-wave singlet pairing $\Tc^d$ is
higher than $\Tc^+$ for $\eta >0.04$.
%
The detailed analysis about the possibility of
this multiple transition will be
given in a subsequent paper.\cite{Fuseya}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Competition between $p$-wave and $d$-wave pair states}
%==============================================================================
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{Tceta.eps}
\caption{Transition temperature for $p$($d$)-wave singlet pairing,
$\Tc^{p (d)}$, as a function of $\eta$.
}\label{Tceta}\end{center}\end{figure}
As is shown in Fig. \ref{Tceta}, the transition temperature
for $p$-wave singlet $\Tc^p$ exceeds
$\Tc^d$ when $\eta \siml 0.02$.
%
That is to say, in the region away from the QCP,
$d$-wave singlet pairing is realized and $p$-wave singlet pairing
is formed very close to the QCP.
%
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{Vl.eps}
\caption{Frequency dependence of $V_l (\omega _m )$
}\label{Vl}\end{center}\end{figure}
%
The reason why such $p$-wave singlet pairing arises
near the AF-QCP can be understood
as follows.
%
The frequency dependence of $V_l (\textrm{i}\omega_m)$, which
is calculated numerically by eq. (\ref{intVl}), are shown in
Fig. \ref{Vl}.
%
As is also shown in Fig. \ref{Vl},
it should be reasonable to approximate
the frequency dependence of $V_l (\textrm{i} \omega_m)$ as
\begin{eqnarray}
V_l (\textrm{i} \omega_m) \simeq v_l \ln \frac{\omega_0}{\tilde{\eta}
+ |\omega_m |}.
\label{appVl}
\end{eqnarray}
%
Considering the symmetry property of the gap function, eqs. (\ref{even})
and (\ref{odd}),
we can express each $V_l$ in the form:
\begin{eqnarray}
V_p (\textrm{i}\varepsilon_n -\textrm{i}\varepsilon_n')&=&\frac{v_p}{2}
\Biggl( \ln \frac{1}{\tilde{\eta} +|\varepsilon_n -\varepsilon_n'|}
-\ln \frac{1}{\tilde{\eta} +|\varepsilon_n +\varepsilon_n'|}
\Biggr), \\
V_d (\textrm{i}\varepsilon_n -\textrm{i}\varepsilon_n')&=&\frac{v_d}{2}
\Biggl( \ln \frac{1}{\tilde{\eta} +|\varepsilon_n -\varepsilon_n'|}
+\ln \frac{1}{\tilde{\eta} +|\varepsilon_n +\varepsilon_n'|}
\Biggr).
\end{eqnarray}
%
Let us consider the most singular case, $\varepsilon_n = \varepsilon_n'=\pi T$.
%
Of course this example does not reproduce the present situation exactly,
but gives a simple picture of the competition of the pairing states.
%
Each $V_l$ is given by
\begin{eqnarray}
V_p (0) &=& \frac{v_p}{2}\ln \frac{\tilde{\eta}+2\pi T}{\tilde{\eta}}, \\
V_d (0) &=& \frac{v_d}{2}\ln \frac{1}{\tilde{\eta}(\tilde{\eta}+2\pi T)}.
\end{eqnarray}
%
When the system locates away from the QCP, i.e., $\tilde{\eta}\gg 2\pi T$,
\begin{eqnarray}
V_p (0) &\simeq& \frac{v_p}{2}\frac{2\pi T}{\tilde{\eta}}, \\
V_d (0) &\simeq& \frac{v_d}{2}\Biggl( -2\ln \tilde{\eta}
-\frac{2\pi T}{\tilde{\eta}}\Biggr).
\end{eqnarray}
%
In this case, $V_p $ is almost negligeble compared to $V_d$,
leading to $\Tc^d \gg \Tc^p$.
%
In contrast, when the system locates very close to the QCP, i.e.,
$\tilde{\eta}\ll 2\pi T $,
\begin{eqnarray}
V_p (0) &\simeq& \frac{v_p}{2}\Biggl(
\frac{\tilde{\eta}}{2\pi T}-\ln \frac{\tilde{\eta}}{2\pi T}
\Biggr) \sim -\frac{v_p}{2}\ln \frac{\tilde{\eta}}{2\pi T}, \\
V_d (0) &\simeq& \frac{v_d}{2}\Biggl( -2\ln (2\pi T)-\frac{\tilde{\eta}}{2\pi T}
-\ln \frac{\tilde{\eta}}{2\pi T}
\Biggr) \sim -\frac{v_d}{2}\ln \frac{\tilde{\eta}}{2\pi T}.
\end{eqnarray}
This time $V_p$ is comparable to $V_d$, and in the case of
$|v_p| > |v_d|$, $\Tc^p$ becomes higher than $\Tc^d$.
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{vlkF.eps}
\caption{$k_F$-dependence of $v_l $.
The vertical dotted-line indicates $k_F =\pi/\sqrt{2}$ where
$|\textbf{Q}|$ is equivalent to the diameter of the Fermi surface.
}\label{Vl0}\end{center}\end{figure}
As shown in Fig. \ref{Vl0}, $|v_l|$ considerably depends on $k_F$.
%
In other words,
it is determined by the relation between the Fermi surface
and AF ordering vector $\textbf{Q}$.
%
In the present case,
the most singular pair scattering with $\textbf{q}=\textbf{Q}=
(\pm \pi, \pm \pi)$
uses the points near $(\pm k_F /\sqrt{2}, \pm k_F /\sqrt{2})$, where
the nodes of the $d$-wave pairing exists while that of the
$p$-wave pairing does not. (See Fig. \ref{pdH} (a), (b))
%
Therefore the effects of the singular pair scattering are
suppressed in the case of $d$-wave pairing.
%
This causes the relation $|v_p |>|v_d |$ in the present case.
%
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=1.0\linewidth]{pdH.eps}
\caption{Illustration of the relation between the AF ordering vector
$\textbf{Q}$ and the position of the nodes.
(a) $p$-wave pairing case:
the most singlular pair scattering with $textbf{Q}$
is not suppressed by the existence of the nodes.
(b) $d$-wave pairing case: the scattering with $\textbf{Q}$ is suppressed
by the existence of the nodes.
(c) $d$-wave pairing with FS nested with $\textbf{Q}$ case:
the suppression of pair scattering becomes small.
}\label{pdH}\end{center}\end{figure}
%
Considering the above discussion, we can extend the condition
of the emergence of the $p$-wave singlet pairing to the general case
as following.
%
Firstly, the pair scattering interaction must have a sharp peak in
frequency whose width is smaller than the temperature (in the present
case, $\eta \ll T$).
%
Secondly, the most singular pair scattering with vector $\textbf{Q}$
has to be canceled out by the nodes of the $d$-wave pairing.
%
In the case of Hubbard model at the half-filling, for example,
the Fermi surface is nested with $\textbf{Q}=(\pm \pi, \pm \pi)$
and the suppression of pair scattering becomes small,
namely, $|v_p | \sim |v_d |$ (see Fig \ref{pdH} (c)).
%
So, it would be difficult for the $p$-wave singlet pairing
to prevail over the $d$-wave singlet pairing on the Hubbard model,
even near the AF-QCP.
%
How about the cases of CeCu$_2$Si$_2$ and CeRhIn$_5$?
%
The $1/T_1$ of both compounds rapidly increase approaching
the AF+SC phase from SC phase.\cite{YKawasaki1,SKawasaki}
%
This indicates that the system is located very near the QCP,
which would satisfy the first condition mentioned above.
%
In CeRhIn$_5$, the most singular pair scattering vector is
$\textbf{Q}=(\pi, \pi, 0.297\pi)$\cite{Bao}.
%
Its Fermi surface does not nest in the direction $\textbf{Q}$
and is satisfied the relation
$\textbf{k}_F - \textbf{k}'_F =\textbf{Q}$
in the direction $(\pm 1, \pm 1, 0)$.
\cite{Settai}
%
The feature that $\textbf{Q}$ is independent from the nesting vector
can be understood from the dual nature of electron.
%
According to the itinerant-localized duality model\cite{Kuramoto},
the spin susceptibility $\chi (q, \textrm{i}\omega_m)$ is
expressed in terms of the polarization function
$\Pi (q, \textrm{i}\omega_m)$ of quasiparticles
and the exchange interaction $J(q, \textrm{i}\omega_m)$ between
localized component of electrons as follows:
\begin{eqnarray}
\chi (q, \textrm{i}\omega_m)^{-1}
=\chi_0 (\textrm{i}\omega_m)^{-1} -2\lambda^2 \Pi (q, \textrm{i}\omega_m)
-J(q, \textrm{i}\omega_m),
\end{eqnarray}
where $\chi_0$ is the local susceptibility expressing an effect of
Kondo-like correlation, and $\lambda$ is the renormalized
spin-fermion coupling.
%
Even if the Fermi surface is not nested, i.e., $\Pi $ is not enhanced,
the AF ordering can be triggerd by $J(\textbf{Q})$.
%
Therefore, CeRhIn$_5$ would satisfy both conditions.
%
In the case of CeCu$_2$Si$_2$, a situation is much more complicated
($\textbf{Q}\sim (0.27\pi, 0.27\pi, 0.52\pi)$\cite{Knebel}),
but its Fermi surface seems to satisfy the condition.\cite{Harima}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Gaplessness of odd-frequency pairing}
%==============================================================================
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{gap.eps}
\caption{Frequency dependence of the gap function of the $p$-wave singlet
pairing at $\Tc^+ /\varepsilon _F = 7.0 \times 10^{-2}$
and $\Tc^- /\varepsilon _F = 1.0 \times 10^{-4}$for $\eta=0.04$.
The solid line indicates the results of Pad\'e approximation.
}\label{gap}
\end{center}
\end{figure}
The frequency dependence of the gap function of the $p$-wave singlet
pairing at $\Tc^+ = 7.0 \times 10^{-2}$ and $\Tc^-
= 1.0 \times 10^{-4}$ for $\eta=0.04$
are shown in Fig. \ref{gap}.
%
For $T=\Tc^+$, at first sight, the gap seems to be divergent
when $\varepsilon _n$ approaches zero.
%
The gap at $T=\Tc^-$, on the other hand, has a peak at $\varepsilon_n^* =
6.19 \times 10^{-2}$, which is comparable to $\Tc^+ $,
and vanishes for $\varepsilon \to 0$, i.e.,
it exhibits a gapless-like structure.
%
This feature seems robust against variations of
$\eta , k_F$ or $g$.
%
Therefore, we can suspect that the peak and the gapless-like structure is
only hidden for $T=\Tc^+$.
%
At the present stage, however, we cannot declare wheather the present
$p$-wave singlet gap function is completely gapless or not.
%
So, we have to perform the analytic continuation of the gap function.
%
In order to do that, first we apply the Pad\'e approximation
to the calculated gap function.
%
Since the gap function is expected to be proportional to
$|\varepsilon_n |^{-2}$
for large $\varepsilon_n $,
we take the form:
\begin{eqnarray}
\Delta_{\textrm{P}} (\textrm{i}\varepsilon_n )
=\frac{a_5 (\varepsilon_n + a_1 )}{1+ a_2 \varepsilon_n
+a_3 \varepsilon_n^2 + a_4 \varepsilon_n^3},
\end{eqnarray}
for $\varepsilon _n >0$.
%
We apply the least squares method to the calculated gap function
$\Delta (\textrm{i}\varepsilon_n )$by
$\Delta_{\textrm{P}}(\textrm{i}\varepsilon_n )$ and obtain
$a_1 = -1.2455 \times 10^{-5}, a_2 =2.2966\times 10,
a_3 = 1.6980 \times 10^2 , a_4 =7.7980 \times 10^2 ,
a_5 = 6.6046 \times 10$.
%
Using these values of $a_{1\sim 5}$, we display
$\Delta_{\textrm{P}}(\textrm{i}\varepsilon_n )$ in Fig. \ref{gap},
in which $\Delta_{\textrm{P}}(\textrm{i}\varepsilon_n )$
shows a good agreement with $\Delta (\textrm{i}\varepsilon_n )$.
%
The analytic continuation of this form becomes
\begin{eqnarray}
\textrm{Re} \Delta_{\textrm{P}}(\varepsilon )&=&
\frac{a_1 a_5 (1-a_3 \varepsilon^2 )+a_5 \varepsilon^2
(a_2-a_4 \varepsilon^2 )}{(1-a_3 \varepsilon^2 )^2
+\varepsilon^2 (a_2 -a_4 \varepsilon^2)}, \label{PadeRe}\\
\textrm{Im} \Delta_{\textrm{P}}(\varepsilon )&=&
\frac{a_5 \varepsilon (a_3 \varepsilon^2 -1)+a_1 a_5 \varepsilon
(a_2-a_4 \varepsilon^2 )}{(1-a_3 \varepsilon^2 )^2
+\varepsilon^2 (a_2 -a_4 \varepsilon^2)}. \label{PadeIm}
\end{eqnarray}
%
From eqs. (\ref{PadeRe}) and (\ref{PadeIm}), the gap function
at $\varepsilon = 0$ is reduced to
$\textrm{Re} \Delta_{\textrm{P}}(0)=-8.2260 \times 10^{-4}, \,\,
\textrm{Im} \Delta_{\textrm{P}}(0)=0$.
%
Therefore, we can conclude the present $p$-wave singlet superconductivity
is a technically gapless one.
%
We can determine the maximum of the gap, $\Delta_0 (T)$ below $\Tc$,
by assuming that the gap has the same frequency dependence
as that of the eigenfunction at $T \ll \Tc$, where the eigenvalue $\lambda$
exceeds unity further.
%
Namely,
$\Delta (\textrm{i}\varepsilon_n ; T) \equiv \Delta_0 (T)
\Delta_P (\textrm{i}\varepsilon_n ; T\ll\Tc)$.
%
At T=0, we obtained $\Delta_0 (0) / \Tc = 0.42$ (by setting parameters
as $a_1 = -1.5256 \times 10^{-5}, a_2 = 4.6835 \times 10,
a_3 = 6.4780 \times 10^2, a_4 = 2.8209 \times 10^3,
a_5 = 1.1492 \times 10^2$, which corresponds to $\eta = 0.01$ case),
which is much smaller
than the BCS relation $\Delta_0 (0) / \Tc = 1.765$.
%
We can also discuss the gapless nature in more general form as follows.
%
The linearized gap equation for the odd-frequency gap
can be written as
\begin{eqnarray}
\Delta_p (\textrm{i}\varepsilon_n ) &=&-N_F \int_{\infty}^{\infty}
\textrm{d}\xi T\sum_{\varepsilon_n'}
\frac{V(\textrm{i}\varepsilon_n -\textrm{i}\varepsilon_n ')}
{|\varepsilon_n'|^2 +\xi ^2}
\frac{1}{2}\bigl[ \Delta_p (\textrm{i}\varepsilon_n' )
-\Delta_p (-\textrm{i}\varepsilon_n' ) \bigr] \\
&=&-N_F \textrm{P}\int_{\infty}^{\infty}
\frac{\textrm{d}x}{8}
\coth \frac{x}{2T}\frac{1}{x+\textrm{i}\varepsilon_n}
\bigl[ V(-x-\textrm{i}\delta )-V(-x+\textrm{i}\delta ) \bigr]
\times
\bigl[ \Delta (\textrm{i}\varepsilon_n +x)-
\Delta (-\textrm{i}\varepsilon_n -x)
\bigr] \nonumber \\
& &-N_F \textrm{P}\int_{\infty}^{\infty}
\frac{\textrm{d}x}{8}
\tanh \frac{x}{2T}\frac{1}{x+\textrm{i}\delta}
\bigl[ V(\textrm{i}\varepsilon_n -x)-V(\textrm{i}\varepsilon_n +x) \bigr]
\times
\bigl[ \Delta (x+\textrm{i}\delta)-\Delta (-x-\textrm{i}\delta )
\bigr].
\label{gapeqM}
\end{eqnarray}
%
The analytic continuation, $\textrm{i} \varepsilon_n \to
\varepsilon + \textrm{i}\delta$
in eq. (\ref{gapeqM}), results in the following gap equation:
\begin{eqnarray}
\Delta (\varepsilon +\textrm{i}\delta )
&=&-N_F \textrm{P}\int_{\infty}^{\infty}
\frac{\textrm{d}x}{8}
\Bigl\{
\coth \frac{x-\varepsilon }{2T}
\bigl[ V(\varepsilon -x -\textrm{i}\delta )-
V(\varepsilon -x +\textrm{i}\delta )
\bigr] \nonumber \\
& &\qquad \qquad \qquad \qquad +\tanh \frac{x}{2T}
\bigl[ V(\varepsilon -x +\textrm{i}\delta )
-V(\varepsilon +x +\textrm{i}\delta )
\bigr] \Bigr\} \nonumber \\
& &\qquad \qquad \qquad
\times \frac{\Delta (x+\textrm{i}\delta )-\Delta (-x-\textrm{i}\delta )}
{x+\textrm{i}\delta}.
\label{anagap}
\end{eqnarray}
The analytic continuation of the effective pairing interaction
$V_l (\textrm{i}\omega_m )$, (\ref{appVl}), becomes
\begin{eqnarray}
V(\omega +\textrm{i}\delta )=v_l
\bigl( \ln \frac{\omega_0}{\sqrt{\omega ^2 +\tilde{\eta}^2}}
+\textrm{i} \tan ^{-1} \frac{\omega }{\tilde{\eta}}
\bigr)
\label{anaVl}
\end{eqnarray}
%
Substituting eq. (\ref{anaVl}) into eq. (\ref{anagap}),
we obtain
\begin{eqnarray}
\Delta (\varepsilon +\textrm{i}\delta ) &=&
-v_l N_F \textrm{P} \int_{-\infty}^{\infty}
\frac{\textrm{d}x}{8}
\Bigl\{
\tanh \frac{x}{2T}\ln\sqrt{\frac{(\varepsilon +x )^2 +\tilde{\eta}^2}
{(\varepsilon -x )^2 +\tilde{\eta}^2}} \nonumber\\
& &+\textrm{i} \bigl[
2\coth \frac{x-\varepsilon }{2T}
\tan ^{-1} \frac{\varepsilon -x}{\tilde{\eta}}
+\tanh \frac{x}{2T}
\bigl(
\tan ^{-1} \frac{\varepsilon -x}{\tilde{\eta}}
-\tan ^{-1} \frac{\varepsilon +x}{\tilde{\eta}}
\bigr)
\bigr] \Bigr\} \nonumber \\
& &\qquad \qquad \qquad \qquad \qquad
\times \frac{\Delta (x+\textrm{i}\delta )-\Delta (-x-\textrm{i}\delta )}
{x+\textrm{i}\delta}.
\label{anagapeq}
\end{eqnarray}
%
Let us express the real (imaginary) part of the gap as
$\Delta '(\varepsilon )$ ($\Delta '' (\varepsilon )$),
i.e., $\Delta (\varepsilon +\textrm{i}\delta ) \equiv
\Delta '(\varepsilon ) +\textrm{i}\Delta ''(\varepsilon )$.
%
Then, each part of (\ref{anagapeq}) are reduced to
\begin{eqnarray}
\Delta '(\varepsilon)
&=&-v_l N_F \textrm{P}\int_{-\infty}^{\infty}
\frac{\textrm{d}x}{8}
\Biggl\{
\tanh \frac{x}{2T} \ln\sqrt{\frac{(\varepsilon +x )^2 +\tilde{\eta}^2}
{(\varepsilon -x )^2 +\tilde{\eta}^2}}
\frac{\Delta '(x)-\Delta '(-x)}{x} \nonumber \\
%
& &-\Bigl[
2\coth \frac{\varepsilon -x}{2T} \tan ^{-1}
\frac{\varepsilon -x}{\tilde{\eta}}
+\tanh \frac{x}{2T}
\bigl(
\tan ^{-1} \frac{\varepsilon -x}{\tilde{\eta}}
-\tan ^{-1} \frac{\varepsilon +x}{\tilde{\eta}}
\bigr)
\Bigr]
\frac{\Delta ''(x)+\Delta ''(-x)}{x}
\Biggr\} \nonumber \\
\\
%
\Delta '' (\varepsilon)
&=&-v_l N_F \textrm{P}\int_{-\infty}^{\infty}
\frac{\textrm{d}x}{8}
\Biggl\{
\tanh \frac{x}{2T} \ln\sqrt{\frac{(\varepsilon +x )^2 +\tilde{\eta}^2}
{(\varepsilon -x )^2 +\tilde{\eta}^2}}
\frac{\Delta ''(x)+\Delta ''(-x)}{x} \nonumber \\
%
& &-\Bigl[
2\coth \frac{\varepsilon -x}{2T} \tan ^{-1}
\frac{\varepsilon -x}{\tilde{\eta}}
+\tanh \frac{x}{2T}
\bigl(
\tan ^{-1} \frac{\varepsilon -x}{\tilde{\eta}}
-\tan ^{-1} \frac{\varepsilon +x}{\tilde{\eta}}
\bigr)
\Bigr] \nonumber \\
%
& & \times \biggl[ \frac{\Delta '(x)-\Delta '(-x)}{x}
+\bigl\{ \Delta ''(x) +\Delta ''(-x) \bigr\} \pi \delta (x)
\biggr]\Biggr\}.
\end{eqnarray}
Note here that we can show easily
$\Delta '(\varepsilon )=-\Delta '(-\varepsilon)$,
$\Delta (\varepsilon )=\Delta ''(-\varepsilon)$.
%
At $\varepsilon =0$,
\begin{eqnarray}
\Delta '(0) &=&0, \label{Delta'0}\\
\Delta ''(0) &=& -v_l N_F \textrm{P}\int_{-\infty}^{\infty}
\frac{\textrm{d}x}{2}
\biggl( \coth \frac{x}{2T}-\tanh \frac{x}{2T}
\biggr)
\tan ^{-1}\frac{x}{\tilde{\eta}} \cdot
\frac{\Delta '(x)}{x}
-v_l N_F \pi \frac{T}{\tilde{\eta}}\Delta ''(0).
\label{Delta''0}
\end{eqnarray}
%
From eq. (\ref{Delta''0}),
\begin{eqnarray}
\biggl(1+v_l N_F \pi \frac{T}{\tilde{\eta}}
\biggr) \Delta ''(0) = -v_l N_F \textrm{P}\int_{-\infty}^{\infty}
\frac{\textrm{d}x}{2}
\biggl( \coth \frac{x}{2T}-\tanh \frac{x}{2T}
\biggr)
\tan ^{-1}\frac{x}{\tilde{\eta}} \cdot
\frac{\Delta '(x)}{x}.
\label{Delta''02}
\end{eqnarray}
%
The right hand side of eq. (\ref{Delta''02}) is constant,
and $1+v_l N_F \pi T/\tilde{\eta} \to -\infty $
when $\tilde{\eta}\to 0$.
%
Therefore, in order to satisfy eq. (\ref{Delta''02}),
\begin{eqnarray}
\Delta ''(0) \to 0 \hspace{1cm} \textrm{for}
\hspace{1cm} \tilde{\eta}\to 0
\label{Delta''03}
\end{eqnarray}
is needed.
%
From eqs. (\ref{Delta'0}) and (\ref{Delta''03}),
we reach the fact that the present $p$-wave
singlet gap function leads to an essentially gapless superconductivity
at the QCP.
%
Considering the low-frequency structure of the gap function,
we assume that the gap function can be approximate in the form:
\begin{eqnarray}
\Delta (\textbf{k}, \textrm{i}\omega_m )= \frac{\Delta_0}{\Tc}
\textrm{i}\omega_m \phi_p (\textbf{k}).
\end{eqnarray}
%
Then we find from the poles of Green's function that
\begin{eqnarray}
E_{\textbf{k}}=\frac{\xi}{\sqrt{1+(\Delta_0 /\Tc)^2 \phi_p^2
(\textbf{k})}}.
\end{eqnarray}
%
Therefore, the quasiparticle spectra with such gap function is
gapless, i.e., there is no difference in excitations between
the nomal states and the present SC states,
except for the effective mass enhancement:
\begin{eqnarray}
m_{\textbf{k}}^* = m\sqrt{1+(\Delta_0 /\Tc)^2 \phi_p^2 (\textbf{k})}.
\end{eqnarray}
%
This mass enhancement is, however, very small
due to the smallness of $\Delta_0 /\Tc \sim 0.42$ as mentioned above,
so that the specific heat or the NMR relaxation rate does not
show any significant change at $\Tc$.
%
This would correspond to the $1/T_1$ behavior
observed in Ce$_{0.99}$Cu$_{2.02}$Si$_2$
\cite{Ishida,YKawasaki1,SO5}
which shows almost the same behavior as the normal Fermi liquid state.
%
If the $1/T_1 \propto T$ behavior is due to the impurity scattering,
$1/T_1 $ should shows a significant reduction at $\Tc$ and exhibit
$1/T_1 \propto T$ well below $\Tc$. \cite{Sr2RuO4}
%
This is the reason why we conclude the $1/T_1 \propto T$
is not due to the impurity
scattering but due to the odd-frequency gap.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Coexistence of AF and SC order}
%==============================================================================
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{VlAF.eps}
\caption{Illustration of the frequency dependence of $V_l (\textrm{i}
\omega_m ) $in the AF backgrounds,
where $\omega_m^\pm \equiv \varepsilon_n \pm \varepsilon_n '$.
The $p$-wave channel dominates the $d$-wave channel
because $v_p > v_d $ and the unsaturated behavior of
$V_l (\textrm{i}\omega_m ) $.
}\label{VlAF}
\end{center}
\end{figure}
In the previous sections, we discussed the emergence of the gapless
$p$-wave singlet pairing prevailing the $d$-wave singlet
pairing in the paramagnetic (PM) backgrounds.
%
The experimental results \cite{Ishida,YKawasaki1,SO5,YKawasaki2,
Mito,SKawasaki,Kitaoka1,Fisher} suggest that the gapless
superconductivity is realized in the AF backgrounds
rather than in the PM backgrounds.
%
(In the SC phase very close to the phase boundary, the present
theory predicts that the gapless $p$-wave singlet superconductivity
is realized, but the detailed experimental results in this region
have not been obtained yet.)
%
Here we discuss that the gapless $p$-wave singlet can also be realized
in rather wide region in AF+SC phase.
%
Since the transverse spin susceptibility $\chi _{\perp}$
would be much dominant than the longitudinal one $\chi _{\parallel }$,
even if the damping effect would have been taken into account,
the pairing interaction in the AF backgrounds may given as follows:
\begin{eqnarray}
V(\textbf{q}, \textrm{i}\omega_m ) =
g^2 \chi_{\perp}(\textbf{q}, \textrm{i}\omega_m )
\equiv \frac{g^2 N_F}{S^2 \hat{\textbf{q}}^2 +|\omega_m |^2},
\label{AFint}
\end{eqnarray}
where $S$ corresponds to the spin-wave velocity.
%
This interaction is similar to that just at QCP.
%
Namely, the frequency dependence
of $V_l ^{\textrm{AF}}(\textrm{i} \omega _m )$ can be approximated as
\begin{eqnarray}
V_l ^{\textrm{AF}}(\textrm{i} \omega _m )\simeq v_l \ln \frac{\omega_0 '}
{|\omega_m |^2}.
\end{eqnarray}
%
Comparing this with the expression (\ref{appVl}),
the interaction $V_l ^{\textrm{AF}}$ of the AF side
can be obtain by the transformation
$V_l ^{\textrm{AF}} (\textrm{i} \omega _m )\to 2V_l ^{\textrm{PM}}
(\textrm{i} \omega _m ; \tilde{\eta} =0)$.
%
Figure \ref{VlAF} illustrates this situation.
%
There is no saturation in contrast to $V_l ^{\textrm{PM}}$
in the PM side (see Fig.
\ref{Vl} ), which stimulates the emergence of $p$-wave singlet pairing.
%
How the $p$-wave states dominate the $d$-wave one
can be understood in the similar scenario discussed in \S 3,
but this time we present much more generalized and intuitive
picture about how the $p$-wave dominates the $d$-wave.
%
Each $V_l ^{\textrm{AF}}$ can be expressed in the form
\begin{eqnarray}
V_p ^{\textrm{AF}}(\textrm{i} \omega _m ) &\simeq &
v_p \Biggl( \ln \frac{1}{\omega_m^-} -\ln \frac{1}{\omega_m^+}
\Biggr) \\
V_d ^{\textrm{AF}}(\textrm{i} \omega _m ) &\simeq &
v_d \Biggl( \ln \frac{1}{\omega_m^-} +\ln \frac{1}{\omega_m^+}
\Biggr),
\end{eqnarray}
where $\omega_m^{\pm } \equiv \varepsilon_n \pm \varepsilon_n '$.
%
From these expressions, we can easily find that
$V_d ^{\textrm{AF}}>V_p ^{\textrm{AF}}$, in general, for $v_d \sim v_p$
because the first term and the second term are added for
$d$-wave, while they offset each other for $p$-wave.
%
In the case $v_p$ is moderately larger than
$v_d $ as displayed in Fig. \ref{VlAF},
however, the first term of $V_p ^{\textrm{AF}}$ is much larger
than $V_d ^{\textrm{AF}}$, so that
the $p$-wave singlet pairing dominates the $d$-wave one.
%
This situation would be realized easier in the AF side
rather than that in the PM side because of the factor 2 and
the unsaturation behavior of $V_l ^{\textrm{AF}}$.
%
Judging from above, no matter how the detail is,
the $p$-wave pairing prevails $d$-wave one when the frequency
dependence of the pairing interaction
takes the form as shown in Fig. \ref{VlAF}.
%
\begin{figure}[btp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.5\linewidth]{hotspot.eps}
\caption{Missing Fermi surface due to the AF energy gap and
the definition of $\varphi$.
}\label{hotspot}
\end{center}
\end{figure}
%
Below $T_{\textrm{N}}$, however,
the energy gap is formed around the hot spot
(Fig. \ref{hotspot}) and the missing part of the Fermi surface
spreads out as the AF ordering is developed.
%
It leads to the remarkable reduction of $\Tc$ because
the dominant pair scattering is suppressed due to the existence of
the AF gap.
%
To simulate this situation, we parameterize the
effect of the AF order as the
angle $\varphi$ of the missing part of the Fermi surface as displayed
in Fig. \ref{hotspot}.
%
The growth of the AF gap,
spreading of $\varphi$, deprives the pairing interaction
of the most singular scattering with
the AF ordering vector $\textrm{Q}$,
and suppresses $V_l ^{\textrm{AF}}$.
%
It is noted that this suppression is not the saturation effect displayed
in Fig. \ref{Vl}, but the reduction of $v_l$.
%
This reduction of $v_l$ is remarkable for $v_p$, because
the enhancement of $v_p$ is mainly due to the scattering
using the hot spots.
%
We can also discuss the gapless nature as discussed in \S 4.
%
In the AF backgrounds, the pairing interaction $V_l ^{\textrm{AF}}$
is given by the same form as that in the PM backgrounds
with $v_l \to 2v_l$ in the limit $\eta \to 0$.
%
Thus we reach the same conclution that the gap function in the AF side
is always gapless:
\begin{eqnarray}
\Delta _{\textrm{AF}}' (0) &=& 0, \\
\Delta _{\textrm{AF}}'' (0) &=& 0.
\end{eqnarray}
%
Together with the $\Tc$'s in the PM phase,
the calculated $\Tc$'s near the QCP are shown in Fig. \ref{etapsi}
%
Here we set $S=1.2$ and the other parameter $g^2 N_F$
and $k_F$ to be the same as those in the PM backgrounds.
%
\begin{figure}[tbp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{AFgap.eps}
\caption{Frequency dependence of the gap function of the $p$-wave
singlet pairing in the AF background.
The solid line indicates the results of Pad\'e approximation.
}\label{AFgap}
\end{center}
\end{figure}
%
\begin{figure}[tbp]
%h=here, t=top, b=bottom, p=separate figure page
\begin{center}\leavevmode
\includegraphics[width=0.7\linewidth]{etapsi.eps}
\caption{Transition temperatures near the QCP.
AF+$d$SC: $d$-wave singlet pairing ($d$-S) with line nodes;
AF+$p$SC: gapless $p$-wave singlet ($p$-S);
$p$SC: gapless $p$-S; $d$SC: $d$-S with line nodes.
}\label{etapsi}
\end{center}
\end{figure}
%
In the PM region, the distance from the QCP is
parameterized by $\eta$, and in the AF region, that is parameterized
by $\varphi$.
%
Indeed, the gap in AF backgrounds, shown in Fig. \ref{AFgap},
has a gapless-like structure, and the parameters of Pad\'e approximation
are obtained as $a_1 =-3.3415\times 10^{-3}, a_2 =2.5624 \times 10,
a_3 =-7.7475 \times 10, a_4 = 3.3734 \times 10^2, a_5 =4.8612 \times 10$.
%
The gapless odd-frequency $p$-wave singlet pairing is realized only near
the QCP, and $d$-wave singlet pairing prevails
$p$-wave singlet pairing in the PM region away from the QCP.
%
According to this phase diagram, the observed $1/T_1$ may be explained
as follows.
%
Near the QCP, i.e., AF+$p$SC and $p$SC region displayed
in Fig. \ref{etapsi},
$1/T_1$ does not show any significant reduction due to
the gapless SC state.
%
Away from the QCP (AF+$d$SC and $d$SC region),
$1/T_1 \propto T^3$ behavior is observed.
%
This phase diagram agrees qualitatively with several experiments
in CeCu$_2$Si$_2$ and CeRhIn$_5$.
%
Polycrystalline sample of Ce$_{0.99}$Cu$_{2.02}$Si$_2$ shows gapless
behavior at ambient pressure, and $1/T_1 \propto T^3$ behavior
at pressures $P \simg 0.1$GPa.\cite{Ishida,YKawasaki1}
%
This property can be understood that Ce$_{0.99}$Cu$_{2.02}$Si$_2$
at $P=0$GPa is located in $p$SC phase and at $P \simg 0.1$GPa, $d$SC
phase.
%
CeCu$_2$(Si$_{0.99}$Ge$_{0.01}$)$_2$ compounds shows the AF ordering
at $T<0.75$K, and below 0.5K, the gapless superconductivity,
coexisting with the AF order, at ambient pressure.
%
This compounds also exhibits the line-node gap under the pressure
$P=0.85$GPa.\cite{YKawasaki2}
%
Therefore, this compounds would be located in AF+$p$SC phase at
ambient pressure and in $d$SC phase at $P=0.85$GPa.
%
The existence of $p$SC phase in this sample cannot be recognized
from the present experimental data.
%
Similarly, in CeRhIn$_5$ the pressure region $1.6\siml P \siml 1.75$GPa
would correspond to AF+$p$SC and $P>2.1$GPa
would correspond to $d$SC.\cite{Mito,SKawasaki,Kitaoka1}
%
For each case, the AF+$d$SC phase, where the $d$-wave singlet
SC states and the AF states coexist, have not been observed yet.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Conclusion}
%==============================================================================
In the present paper, we have shown that the $p$-wave singlet
superconductivity with the gap function which is odd in both
momentum and frequency prevails the $d$-wave singlet
superconductivity, at the AF-QCP and in the AF side of AF-QCP.
%
This odd-frequency $p$-wave singlet pairing is stabilized in
the AF states, and is able to coexist with the AF order.
%
The characteristic properties of this $p$-wave
singlet superconductivity is that
i) there is no gap in the quasiparticle spectrum;
ii) $d$-wave pairing arises apart from the AF-QCP even in the AF state;
iii) In some parameter region, $p$-wave pairing is realized only
in the window of temperatures $\Tc^- <T<\Tc^+$.
%
The first and the second property would explain the experiments.
%
Namely, the $1/T_1 \propto T$ behavior is observed in the AF+SC region or
in the boundary of CeCu$_2$Si$_2$ or CeRhIn$_5$,
and $1/T_1 \propto T^3 $ behavior
in the PM side away from the boundary.
%
From the second one, we can predict that
the $d$-wave pairing will be observed again away from the AF-QCP
in the AF states.
%
The condition of the emergence of the $p$-wave singlet pairing
is that the FS is not nested, so that the hot points of surface
connected by the AF wave vector $\textbf{Q}=(\pi, \pi )$ are isolated,
while the AF order is induced by the exchange interaction between
localized component of spin degrees of freedom.
%
Although the structure of the Fermi surface of CeCu$_2$Si$_2$
\cite{Harima} and CeRhIn$_5$\cite{Settai}
is much more complicated than the present model,
they seem to satisfy the condition from inspection of the shape of
the Fermi surface obtained by band structure calculations.
%
We thus believe that the present theory would distill the fundamental
picture of both CeCu$_2$Si$_2$ and CeRhIn$_5$, and obtain the qualitative
understanding of them.
%
In order to get futher understanding, extensive calculation
considering the momentum-dipendence and the practical dispersion
of the quasiparticle are now in progress.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section*{Acknowledgements}
%==============================================================================
Authors would like to thank S. Kawasaki, G.-q.\ Zheng and Y. Kitaoka
for discussion of experimental results.
%
One of the authors (Y. F) is supported by Research Fellowships of the
Japan Society for the Promotion of Science for Young Scientists.
%
This work was supported by a Grant-in-Aid for COE Research (10CE2004)
from Monbu-Kagaku-sho.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{thebibliography}{99}
%==============================================================================
\bibitem{Steglich}
F. Steglich, J. Aarts, C. D. Bredl, W. Lieke, D. Meschede,
W. Franz and H. Sch\"{a}fer: Phys. Rev. Lett. \textbf{43} (1979)1892.
\bibitem{MSV}
K. Miyake, S. Schmitt-Rink and C. M. Varma: Phys. Rev. B \textbf{34}
(1986) 6554.
\bibitem{Scalapino}
D. J. Scalapino, E. Loh, Jr. and J. E. Hirsch:
Phys. Rev. B \textbf{34} (1986) 8190.
\bibitem{Ueda}
K. Ueda, T. Moriya, Y. Takahashi:
in \textit{Electronic Properties and mechanism of High-Tc Superconductors}
(Tsukuba Symposium, 1991) ed. T. Oguchi et al. (North Holland, 1992) p.145.
\bibitem{Monthoux}
P. Monthoux and G. G. Lonzarich:
Phys. Rev. B \textbf{59} (1999) 14598.
\bibitem{Ishida}
K. Ishida, Y. Kawasaki, K. Tabuchi, K. Kashima, Y. Kitaoka, K. Asayama,
C. Geibel and F. Steglich: Phys. Rev. Lett. \textbf{82} (1999) 5353.
\bibitem{YKawasaki1}
Y. Kawasaki, K. Ishida, T. Mito, C. Thessieu, G.-q. Zheng, Y. Kitaoka,
C. Geibel and F. Steglich: Phys. Rev. B \textbf{63} (2001) 140501(R)
\bibitem{SO5}
Y. Kitaoka, K. Ishida, Y. Kawasaki, O. Trovaelli, C. Geibel and
F. Steglich: J. Phys.: Condens. Matter \textbf{13} (2001) L79.
\bibitem{YKawasaki2}
Y. Kawasaki, K. Ishida, K. Obinata, K. Tabuchi, K. Kashima and Y. Kitaoka:
Phys. Rev. B \textbf{66} (2002) 224502.
\bibitem{Koda}
A. Koda, W. Higemoto, R. Kadono, Y. Kawasaki, K. Ishida, Y. Kitaoka,
C. Geibel and F. Steglich: J. Phys. Soc. Jpn. \textbf{71} (2002) 1427.
\bibitem{Thomas}
F. Thomas, C. Ayache, I. A. Fomine, J. Thomasson and C. Geibel:
J. Phys.: Condens. Matter \textbf{8} (1996) L51.
\bibitem{Holmes}
A. T. Holmes, A. Demuer and D. Jaccard: Proceedings of SCES'02,
Acta Phys. Pol. B, in press.
\bibitem{Onishi}
Y. Onishi and K. Miyake: J. Phys. Soc. Jpn. \textbf{69} (2000) 3955.
\bibitem{Yuan}
H. Q. Yuan, M. Deppe, G. Sparn, F. M. Grosche, C. Geibel
and F. Steglich: Proceedings of SCES'02, Acta Phys. Pol. B, in press.
\bibitem{Heeger}
H. Hegger, C. Petrovic, E. G. Moshopoulou, M. F. Hundley,
J. L. Sarrao, Z. Fisk and J. D. Thompson: Phys. Rev. Lett
\textbf{84} (2000) 4986.
\bibitem{Mito}
T. Mito, S. Kawasaki, G.-q. Zheng, Y. Kawasaki, K. Ishida, Y. Kitaoka,
D. Aoki, Y. Haga and Y. \=Onuki: Phys. Rev. B \textbf{63} (2001) 220506(R);
cond-mat/0211576.
\bibitem{SKawasaki}
S. Kawasaki, T. Mito, G.-q. Zheng, C. Thessieu, Y. Kawasaki, K. Ishida,
Y. Kitaoka, T. Muramatsu, T. C. Kobayashi, D. Aoki, S. Araki,
Y. Haga, R. Settai adn Y. \=Onuki: Phys. Rev. B \textbf{65} (2001) 020504;
to be prepared.
\bibitem{Kitaoka1}
Y. Kitaoka, Y. Kawasaki, T. Mito, S. Kawasaki, G.-q. Zheng, K. Ishida,
D. Aoki, Y. Haga, R. Settai, Y. Onuki, C. Geibel and F. Steglich:
J. Phys. Chem. Solids \textbf{63} (2002) 1141.
\bibitem{Fisher}
R. A. Fisher, F. Bouquet, N. E. Phillips, M. F. Hundley, P. G. Pagliuso,
J. L. Sarrao, Z. Fisk and J. D. Thompson: Phys. Rev. B \textbf{65} (2002)
224509.
\bibitem{Sato}
N. K. Sato, N. Aso, K. Miyake, R. Shiina, P. Thalmeier, G. Varelogiannis,
C. Geibel, F. Steglich, P. Flude and T. Komatsubara: Nature
\textbf{410} (2001) 340.
\bibitem{Kuramoto}
Y. Kuramoto and K. Miyake: J. Phys. Soc. Jpn. \textbf{59} (1990) 2381;
Prog. Theor. Phys. Suppl. \textbf{108} (1992) 199;
K. Miyake and Y. Kuramoto: Physica B \textbf{171} (1991) 20.
\bibitem{Tou}
H. Tou, Y. Kitaoka, K. Asayama, C. Geibel, C. Schank and F. Steglich:
J. Phys. Soc. Jpn. \textbf{64} (1995) 725.
\bibitem{Kotliar}
G. Kotliar, E. Abrahams, A. E. Ruckenstein, C. M. Varma, P. B. Littlewood
and S. Schmitt-Rink: Erophys. Lett. \textbf{15} (1991) 655.
\bibitem{Maebashi}
K. Miyake and H. Maebashi: J. Phys. Chem. Solids \textbf{62} (2001) 53.
\bibitem{Narikiyo}
K. Miyake and O. Narikiyo: J. Phys. Soc. Jpn. \textbf{71} (2002) 867.
\bibitem{Balatsky}
A. Balatsky and E. Abrahams: Phys. Rev. B \textbf{45} (1992) 13125.
\bibitem{Berezinskii}
V. L. Berezinskii, JETP Lett. \textbf{20} (1974) 287.
\bibitem{Matsuura}
K. Miyake, T. Matsuura and H. Jichu: Prog. Theor. Phys. \textbf{72}
(1984) 652.
\bibitem{Kitaoka2}
Y. Kitaoka, K. Ueda, T. Kohara and K. Asayama: Solid State Commun.
\textbf{51} (1984) 461.
\bibitem{Abrahams}
E. Abrahams, A. Balatsky, J. R. Schrieffer and P. B. Allen:
Phys. Rev. B \textbf{47} (1993) 513.
\bibitem{private}
K. Miyake.: unpublished.
\bibitem{Bao}
W. Bao, P. G. Pagliuso, J. L. Sarrao, J. D. Thompson, Z. Fisk,
J. W. Lynn and R. W. Erwin: Phys. Rev. B \textbf{62} (2000) R14621.
\bibitem{Fuseya}
Y. Fuseya, H. Kohno and K. Miyake: in preperation.
\bibitem{Settai}
R. Settai, H. Shishido, S. Ikeda, Y. Murakawa, M. Nakashima, D. Aoki,
Y. Haga, H. Harima and \=Onuki: J. Phys.: Condens. Matter \textbf{13}
(2001) L627.
\bibitem{Knebel}
G. Knebel, C. Eggert, D. Engelmann, R. Viana, A. Krimmel, M. Dressel
and A. Loidl: Phys. Rev. B \textbf{53} (1996) 11586.
\bibitem{Harima}
H. Harima and A. Yanase: J. Phys. Soc. Jpn. \textbf{60} (1991) 21.
\bibitem{q1d}
The similar properties that the peak position of the gap is scalled
by $\Tc$ is also seen in the odd-energy gap function in
one-dimensional case as is seen in
Y. Fuseya, H. Kohno and K. Miyake: J. Phys.: Condens. Matter
\textbf{14} (2002) L655.
\bibitem{Sr2RuO4}
For example, K. Ishida, H. Mukuda, Y. Kitaoka, Z. Q. Mao, Y. Mori
and Y. Maeno: Phys. Rev. Lett. \textbf{84} (2000) 5387.
\end{thebibliography}
\end{document}