Nature of the Fulde-Ferrell-Larkin-Ovchinnikov phases at low temperature in 2 dimensions
Extracted source text ยท arXiv cond-mat/0306575
% | " ' | |
%\documentstyle[aps]{revtex}
%\documentstyle[aps,graphicx]{revtex}
%\documentstyle[preprint,aps]{revtex} % for preprint form
\documentstyle[aps,graphicx,prl,multicol]{revtex}
%\documentstyle[aps]{revtex}
%\documentstyle[preprint,aps]{revtex} % for preprint form
\input{epsf.tex}
\begin{document}
\draft
\title{Nature of the Fulde-Ferrell-Larkin-Ovchinnikov phases at low
temperature in 2 dimensions}
\author{C. Mora and R. Combescot }
\address{Laboratoire de Physique Statistique,
Ecole Normale Sup\'erieure*,
24 rue Lhomond, 75231 Paris Cedex 05, France}
\date{Received \today}
\maketitle
\begin{abstract}
We investigate, mostly analytically, the nature of the Fulde-Ferrell-
Larkin-Ovchinnikov superfluid phases for an isotropic two-dimensional
superconductor, which is relevant for the experimental observation of
these phases. We show that there is, in the low temperature range, a
cascade of transitions toward order parameters with ever increasing
complexity.
\end{abstract}
\pacs{PACS numbers : 74.20.Fg, 74.25.Op, 74.70.Kn, 74.81.-g }
\begin{multicols}{2}
The transition to Fulde-Ferrel-Larkin-Ovchinnikov (FFLO) superfluid
phases \cite{ff,larkov} has given rise to much work, but a full
theoretical understanding of these inhomogeneous phases has not yet
been reached. This is of high practical importance since experiment
relies heavily on theory when it needs to test if these phases have
actually been observed. Indeed beside their basic theoretical interest
which, in addition to superconductivity and its coexistence with
magnetism \cite{hbbm}, ranges from ultracold gases \cite{rcffan} to
quark matter as it might occur in the core of neutron stars
\cite{bowers,casal}, FFLO phases are also of very high applied
interest, because they correspond to the superconducting phases which
should appear in superconductors with very high critical fields in the
vicinity of the transition to the normal state. A main strategy to observe
these transitions in superconductors is to get rid of the orbital currents,
responsible for the standard disappearance of the superconducting
phase at low critical fields, and reach the regime of Pauli paramagnetic
limitation. This can be done in clean quasi two-dimensional systems
such as organic compounds made of widely separated conducting
planes, or in high $T_{c}$ compounds. In these cases hopping
between planes is very strongly inhibited. So when a strong magnetic
field is applied parallel to the planes, the resulting orbital currents
perpendicular to the planes are very weak and there is basically no
orbital pair breaking effect which leaves open the possibility of
transitions to FFLO phases at much higher fields. Nevertheless the
small interplane coupling is necessary to suppress phase fluctuations
and produce phase which can be properly described by a mean field
theory. Indeed observations of this transition have been claimed quite
recently in such organic compounds as $ \kappa $-(BEDT-
TTF)$_2$Cu(NCS)$_2$ \cite{nam,singleton} or $ \lambda $-(BETS-
TTF)$_2$GaCl$_4$ or FeCl$_4$ \cite{tanatar,uji}.
Actually when the field is not exactly parallel to the planes there are
small orbital currents and associated vortices appear. The physics of the
evolution from this vortex state to the pure FFLO phase is quite
remarkable, both in the linear \cite{bul,shirai,manalo,hbbm} and in the
non linear regime \cite{klrashi} since it involves a series of vortex states
phases corresponding to Landau levels with increasing quantum
number. Compared to this complex structure the limiting 2D FFLO
phase sounds simple. Indeed the FFLO transition in 2D systems is
believed to be second order and in particular Burkhardt and Rainer
\cite{br} have studied in details the transition to a planar phase, where
the order parameter $\Delta ( {\bf r})$ is a simple $ \cos({\bf q}. {\bf
r})$ at the transition. This phase has been found by Larkin and
Ovchinnikov \cite{larkov} to be the best one in 3D at $T=0$ for a
second order phase transition. And in 3D it is also found to be the
preferred one in the vicinity of the tricritical point and below
\cite{matsuo,buz1,cm}, although in this case the transition turns out to
be first order (except at very low temperature). However it is not clear
that this simple order parameter is always the best one since, as first
explored by Larkin and Ovchinnikov, it is in competition with any
superposition of plane waves, provided that their wavevectors have all
the same modulus.
In this paper we explore the low temperature range in 2D for the pure
FFLO phase transition in the simple isotropic case and show that
surprisingly there is a cascade of the second order transitions toward
order parameters with ever increasing complexity. In principle there is
an infinite number of phases. Therefore the pure FFLO phase display a
structure just as rich as the vortex phases that we mentionned above,
and the interplay between both promises to be of exceptional interest. A
first step in this direction has been made recently by Shimahara
\cite{shima2} who, instead of simple cosine, found a transition toward
a superposition of up to six plane waves when looking for more
complex 'cristalline' structures. We find agreement with his results.
However we will see that cristallinity does not give the proper
understanding of these phases. Instead we show that, when the
temperature is lowered toward $T=0$, the cascade of transitions toward
order parameters with an ever increasing number of plane waves is due
to the singular nature of the $T=0$ limit, which is already appearent in
the $T=0$ wavevector dependence of the second order term in the free
energy expansion in powers of the order parameter.
Our study follows the general framework set up by Larkin and
Ovchinnikov \cite{larkov} and used by Shimahara \cite{shima2} in the
2D case. The free energy difference $\Omega \equiv \Omega _{s} -
\Omega _{n}$ between the superconducting and the normal state is
expanded to fourth order in powers of the order parameter $\Delta ({\bf
r}) = \sum \Delta _{{\bf q}_{i}} \exp(i{\bf q}_{i}. {\bf r})$. The
second order contribution to $\Omega$ is $ N _{0} \sum_{{\bf q}}
\Omega _{2}(q, \bar{ \mu },T) | \Delta _{{\bf q}} |^{2}$, where $ N
_{0}$ is the single spin density of states at the Fermi surface and $
\bar{ \mu } = (\mu _{\uparrow} - \mu_{\downarrow})/2 $ is half the
chemical potential difference between the two spin populations. At the
FFLO transition we have $\Omega _{2}(q, \bar{ \mu },T)=0$.
Naturally the actual transition corresponds to the largest possible $ \bar{
\mu }$ at fixed $T$. It is easily seen \cite{larkov} that the
corresponding wavevector $q$ for the order parameter is obtained by
minimizing:
\begin{eqnarray}
I(\bar{q}, \bar{ \mu },T) = 2 \pi T \: {\rm Im}
\sum_{n=0} ^{ \infty} \frac{1}
{\sqrt{( i \bar{\omega } _{n})^{2}-\bar{ \mu }^{2} \bar{q}^{2}}} -
\frac{1}{i \bar{\omega } _{n}}
\label{eq1}
\end{eqnarray}
with respect to $\bar{q}$, where we have introduced the dimensionless
wavevector $ \bar{q} = q v_{F} / 2 \bar{ \mu } $ and $\bar{\omega
}_{n} \equiv \omega _{n} - i \bar{\mu }$ with $ \omega _{n} = \pi T
(2n+1) $ being the Matsubara frequency. This can be rewritten as an
integral over real frequency. At $T=0$ one finds explicitely $
I(\bar{q}, \bar{ \mu },T) = {\rm Re} \ln ( 1+ \sqrt{1-\bar{q}^{2}}) -
\ln 2$, which is minimum for $ \bar{q} = 1 $ , in agreement with
Shimahara \cite{shima1} and Burkhardt and Rainer \cite{br}. We note
that the location $ \bar{q} = 1 $ of the minimum is a singular point for
$ I $ since $ I = \ln(\bar{q}/2)$ for $ \bar{q} > 1 $, so $ I $ is
continuous while its derivative is discontinuous for $ \bar{q} = 1 $.
For $ T \neq 0$ but low enough temperature, the condition giving the
minimum can be expanded in powers of $t ^{1/2}$ where $ t = T /
\bar{\mu}$ is a reduced temperature. We have found that expanding
$\bar{q}-1$ up to $t^{2}$ gives a result in excellent agreement with a
straight numerical evaluation for $ t < 0.1$. Here (and similarly below)
we do not give the somewhat lengthy full expression \cite{longpaper},
but we omit the last two terms and display only the lowest order
correction which is:
\begin{eqnarray}
\bar{q}-1 = \frac{t}{2} \, \ln \frac{\pi }{2t}
\label{leading2}
\end{eqnarray}
The second order term in $\Omega$ fixes only the modulus, given by
Eq.(2), of the $N$ wavevectors entering the plane wave expansion of
the order parameter. This leaves open all the various possible order
parameters arising from any combinations of these plane waves. The
selected state will correspond to the lowest fourth order term, which is
given \cite{larkov} in general by:
\begin{eqnarray}
\frac{ N _{0}}{2} \sum_{i,j} (2- \delta _{{\bf q}_{i},{\bf q}_{j}}) |
\Delta _{{\bf q}_{i}}| ^{2} | \Delta _{{\bf q}_{j}}| ^{2} J(\alpha
_{{\bf q}_{i},{\bf q}_{j}}) & & \\ \nonumber
+ (1-\delta _{{\bf q}_{i},{\bf q}_{j}}-\delta _{{\bf q}_{i},-{\bf
q}_{j}}) \Delta _{{\bf q}_{i}} \Delta _{-{\bf q}_{i}} \Delta ^{*}
_{{\bf q}_{j}} \Delta ^{*}_{-{\bf q}_{j}} \tilde{J} (\alpha _{{\bf
q}_{i},{\bf q}_{j}})
\label{fourthorder}
\end{eqnarray}
where $\alpha _{{\bf q}_{i},{\bf q}_{j}}$ is the angle between ${\bf
q}_{i}$ and ${\bf q}_{j}$. Just as for the second order term $
J(\alpha)$ and $ \tilde{J} (\alpha)$ can be expressed \cite{longpaper}
as frequency integrals, expanded at low temperature $ t \ll 1$ and
evaluated for the optimal value of $ \bar{q}$ (given by Eq.(2) to
leading order). As for the optimal $ \bar{q}$ we have carried this
expansion up to second order in $t ^{1/2}$. If we restrict ourselves to
leading order as in Eq.(2) we find for example for $ J(\alpha)$:
\begin{eqnarray}
16 t \bar{ \mu }^{2} J(\alpha) = \frac{\pi }{\alpha } \frac{1}
{\cosh ^{2}[(1/4) \ln \frac{\pi }{2t} - \beta ^{2} /2]} & & \\
\nonumber
- \frac{8}{\sqrt{ \pi}(1+ \cos ( \alpha /2)) }
\int_{0} ^{ \infty} \! dv \: \frac{\exp(-v ^{2})}{ v ^{2}+\beta ^{2}}
\label{Jalpha}
\end{eqnarray}
where we have set $ \beta ^{2} = (1-\cos ( \alpha /2)) /t $. The result up
to second order in $t ^{1/2}$ is not much more complicated. A similar
result \cite{longpaper} is found for $ \tilde{J} (\alpha)$.
\begin{figure}[h]
\begin{center}
\includegraphics[width=0.48\textwidth,height=0.3
\textheight]{court-T0.01.eps}
\caption{$J(\alpha)$ for values $0.01$ and $0.05$ of
the reduced temperature $\bar{t} = T/(\bar{q} \bar{\mu}) =
t / \bar{q}$
(with essentially $ \bar{q}\approx 1$ in all this
range). The lines are calculated using our low temperature
expansion for $J(\alpha)$, while
the points correspond to the direct numerical summation over Matsubara
frequencies. The dashed line in the insert (which displays the same
results with different scale) is the asymptotic behaviour for
$J(\alpha)$ in the $T \to 0$ limit Eq.(\ref{Jlargealpha}). }\label{fig1}
\end{center}
\end{figure}
\vspace{-8mm}
Plots of $ J(\alpha)$ for various low temperatures can be seen in Fig.
\ref{fig1}, either obtained from these low temperature expressions or
from exact calculations of $J(\alpha)$ obtained by direct numerical
summation over Matsubara frequencies after analytical angular
averaging.
We see that, at this level of accuracy, both agree remarkably well up to
rather high temperatures which shows that our low temperature
expansion is quite under control. Now Fig.1 shows that $ J(\alpha)$
has at low $T$ a quite remarkably structured behaviour which can be
easily understood from some limiting cases. First let us take the limit $
T \rightarrow 0$ at fixed $ \alpha $. This implies $ \beta ^{2}
\rightarrow \infty$, the first term in Eq.(4) goes to zero and the easy
integration leads to:
\begin{eqnarray}
J(\alpha) = - \frac{1 }{4 \bar{ \mu } ^{2}} \frac{1}{\sin^{2} ( \alpha
/2)}
\label{Jlargealpha}
\end{eqnarray}
On the other hand if at fixed $ T $ we take the limit $ \alpha \rightarrow
0$, we have $ \beta ^{2} \rightarrow 0$. The limiting behaviour of the
integral is $ \pi /(2 \beta ) - \pi ^{1/2}$, the dominant divergent
contributions from the two terms in Eq.(4) cancel out giving:
\begin{eqnarray}
J(0) = \frac{1 }{4 \bar{ \mu } ^{2}} \frac{1}{t}
\label{eq23}
\end{eqnarray}
which goes naturally to infinity for $ T \rightarrow 0$. From these two
limiting situations we can understand that, for most of the $ \alpha $
range$, J(\alpha)$ is negative as it can be seen from Eq.(5) and it goes
to large negative values when $ \alpha $ gets very small. On the other
hand for $ \alpha =0$ or very small $ J(\alpha)$ is positive and very
large, as results from Eq.(6) (surprisingly $ J(\alpha)$ starts first to
increase strongly before going down to very negative values).
We have made the same kind of treatment \cite{longpaper} for
$\tilde{J} (\alpha )$. The limit $ T \rightarrow 0$ with fixed $ \alpha $
leads to $ \bar{ \mu } ^{2} \tilde{J}(\alpha) = - \pi / 4 \sin \alpha $
while in the limit $ \alpha \rightarrow 0$ at fixed $ T $ we find $ \bar{
\mu } ^{2} \tilde{J}(\alpha) = - 1 /4 $. These two cases make
reasonable that $\tilde{J} (\alpha)$ is always negative, as we find.
However, just as for $J( \alpha )$, one finds also a singular behaviour
at small $ \alpha $. While for $ \beta \approx \alpha /(8t) ^{1/2} \gg 1$
and $\bar{ \mu } ^{2} \tilde{J} (\alpha)$ diverges as $ - \pi /4 \alpha
$, it goes to the finite value $ - 1/4$ for $ \alpha =0$. Nevertheless the
divergent behaviour in $ \alpha ^{-1}$ is weaker than the one in $
\alpha ^{-2}$ found for $J( \alpha )$ and similarly $J(0) \gg
|\tilde{J}(0) |$ at low $T$. So $J( \alpha )$ will play the dominant role
and at first we omit $\tilde{J} (\alpha)$ from our considerations.
Let us now consider which is the most favorable order parameter. We
will always find that the fourth order term is positive, so we have to
make it as small as possible in order to minimize the free energy. We
note first that there is unevitably a strongly positive contribution,
proportional to $J(0)$, coming from the N terms ${\bf q}_{i}= {\bf
q}_{j}$ in the first sum of Eq.(3). However their unfavorable effect is
relatively reduced when $N$ increases since there are $ N ^{2}$ terms
in this sum. Since we clearly have to avoid other positive contributions
of this kind, we want to forbid the angle domain $ 0 < \alpha <
\alpha_0$ where we can take $\alpha_0$ as giving $J(\alpha_0) = 0$,
which is very close to the (strongly negative) minimum of $J( \alpha )$
as it can be seen on Fig. 1. So the angle between any two different
wavectors ${\bf q}_{i}$ and $ {\bf q}_{j}$ should be at least
$\alpha_0$. On the other hand it is of interest to have angles close to
$\alpha_0$ since they give strongly negative contributions to $ \Omega
$. Since it is better to have the number of wavevectors $N$ as large as
possible, we come from symmetry reasons to the conclusion that we
have to choose the wavevectors ${\bf q}_{i}$ angularly equally
spaced with the angle between two nearest wavevectors as close as
possible to $\alpha_0$. Now $\alpha_0 \rightarrow 0 $ when $T
\rightarrow 0 $. Therefore we have indeed a cascade of transitions
corresponding to order parameters with an ever increasing number $ N$
of wavevectors when $T \rightarrow 0 $. This singular behaviour is
clearly linked to the fact that it is not possible, in particular for
$J(\alpha)$, to perform an expansion in the limit $T \rightarrow 0 $,
i.e. this limit is singular.
More precisely it is quite a difficult problem to handle this minimization
from scratch, and for example to prove that the minimum $ \Omega $ is
not obtained for a completely disordered situations. However symmetry
makes such a result quite unlikely. On the other hand if we take for
granted that the wavevectors are equally spaced angularly it is easy to
prove that the weight $ | \Delta _{{\bf q}_{i}} |$ are all equal, because
this problem can be mapped on a tight binding Hamiltonian on a ring.
Conversely if we assume that all these weights are equal the problem is
equivalent to find the equilibrium position of atoms on a ring with
repulsive short range interaction and attractive long range interaction.
We expect the equilibrium to be a crystalline structure corresponding in
our case to equally spaced wavevectors.
\vspace{-4mm}
\begin{figure}[h]
\begin{center}
\includegraphics[width=0.48\textwidth,height=0.3
\textheight]{alter-comparison.eps}
\includegraphics[width=0.48\textwidth,height=0.3
\textheight]{alter-cascade-fin.eps2}
\caption{G$_2 ^{-1}$ as a function of temperature for various plane
waves number $N$ (for clarity only solutions up to $N=14$ are
presented). The arrows give the values of the successive transitions. In
the upper panel the dashed lines are the exact Matsubara summations.
The full lines in the upper panel, as well as all the lines in the lower
panel, are from the low temperature expansion for $J(\alpha)$ and
$\tilde{J} (\alpha )$. In the insert the full dots are the exact temperature
locations of the transitions together with the corresponding values of
$N$ (the empty dots just give $N-2$). The full line gives the low
temperature evaluation of these transitions from the solution of a simple
transcendental equation, and the almost undistinguishable dashed line
comes from the explicit first iteration solution of this
equation.}\label{fig3}
\end{center}
\end{figure}
\vspace{-8mm}
In the low temperature limit ($t \ll 1$), we can easily be more
quantitative and write simple analytical answers for our problem.
Indeed the result is dominated by the small $ \alpha $ behaviour of
$J(\alpha)$. In this range Eq.(4) becomes:
\begin{equation}\label{asympto}
\bar{\mu}^2 J(\alpha) \simeq
\frac{\pi}{16 t \alpha \cosh^{2} X} - \frac{1}{\alpha^2}
\end{equation}
where $X= x - (1/4) \ln( \pi /2t)$ with $x=\alpha ^{2}/16 t$. The
second term is the asymptotic form Eq.(\ref{Jlargealpha}). The zero of
$J(\alpha)$ we are looking for is found for large $X$ and satisfies
$e^{x} =(k/ \sqrt{2}) x^{1/4} $ with $k = (2 \pi ^{3}/ t ^{2})
^{1/4}$, which is easily solved iteratively for large $k$. The leading
order $ x = \ln (k / \sqrt{2}) $ gives $ \alpha_0 = ( 8t \ln [(\pi
^{3}/2)^{1/2}/t] ) ^{1/2}$. The first order iteration is $ x = \ln (k /
\sqrt{2}) + 1/4 \ln (\ln (k / \sqrt{2}) $. Then the optimum value of $N$
is given by the integer value of $ 2 \pi /\alpha_0$, which conversely
fixes the temperature at which the system switches from $N-2$ to $N$
wavevectors. In the insert of Fig.3, one sees that the first iteration gives
results almost undistinguishable from the exact solution of the above
transcendental equation for $x$.
Exact results can be obtained by handling numerically the various steps.
The minimization of the free energy with respect to $ | \Delta _{{\bf
q}} |$ gives $\Omega / N_0 = - \Omega_2 ^2 (q,\bar{\mu},T) /
G_2(N)$ with:
\begin{equation}
\label{G2odd}
N G_2(N) = 2 J(0) + 4 \sum_{n =1} ^{N-1} J(2 \pi n / N)
\end{equation}
where we have omitted the $\tilde{J}$ terms for simplicity. One sees in
Fig. 3 the results (including the $\tilde{J}$ terms) for $ G_2^{-1}
(N)$, i.e. essentially the free energy, for various values of $N$. The
stable state has the highest $ G_2^{-1} (N)$. The cascade of phase
transitions is clearly seen in the lower panel, corresponding to the lower
temperature range. The arrows show the successive transition
temperatures. The results from straight numerical Matsubara summation
can not be distinguished from those obtained from our low $T$
analytical expression for $J(\alpha)$ and $\tilde{J}(\alpha)$, except for
the higher temperature, as seen in the upper panel. The critical
temperatures for switching from $N-2$ to $N$ are shown as dots in the
insert of Fig. 3. It can be seen that our above asymptotic calculation for
these temperatures, shown as full curves, are actually quite good.
At low temperature we can make an asymptotic calculation of $ G_2
(N)$. As seen on Fig. 1 $J(\alpha)$ switches rapidly above $\alpha_0 $
to its large angle asymptotic behaviour. Moreover at low $T$ the
number of plane waves N is large and in Eq.(\ref{G2odd}) the sum is
dominated by the small angles terms, which leads \cite{longpaper} to:
\begin{eqnarray}\label{G21}
\bar{\mu}^2 G_2(N) &\simeq& \frac{1}{2 N t} - \frac{N}{3}
\label{G22}
\end{eqnarray}
This expression corresponds in Fig. 3 to the rising part, on the low
temperature side, of $ G_2^{-1} (N)$. On the other hand the downturn
is due to contributions from the positive part of $J(\alpha)$ which can
not be evaluated so easily. When we substitute in Eq.(\ref{G22}) the
value $N = 2 \pi / \alpha _0$ for the optimum plane wave number we
have:
\begin{equation}
\bar{\mu}^2 G_2(N) = \frac{\alpha_0}{4 \pi t} \left( 1-
\frac{8 \pi^2 t}{3 \alpha_0^2} \right)
\end{equation}
Since $ 8t / \alpha_0^2 \sim 1 / \ln(1/t)$ from our above evaluation, we
find that $ G_2(N)$ is always be positive in the low temperature range,
which means that the transition stays always second order.
When we take into account the contribution of the $\tilde{J}$ terms in
Eq. 3 that we had omitted so far, one gets an additional $- \ln N$
contribution in Eq.(\ref{G21}), which arises only when $N$ is even
because for equally spaced wavevectors, the $\tilde{J}$ terms
contribute only in this case since they require opposite wavevectors.
This contribution is just enough to systematically tilt the balance in
favor of even $N$. In this respect the phase $N=3$ found by
Shimahara \cite{shima2} appears as an exception. Otherwise our
cascade of phase transitions contains only order parameters with an
even number of plane waves, and the order parameter is just a real sum
of cosines, with still a degeneracy due to the freedom of choosing the
phases in these cosines.
We note that, although the transitions between the various FFLO phases
and the normal phase are second order, we expect the transitions
between different FFLO phases in the superfluid domain to be first
order since one can not go continuously from a given order parameter to
the next one. Finally we have seen that these successive transitions are
directly due to the singularity which occurs in 2 D at $T=0$. This
singularity itself arises because the two Fermi circles corresponding to
opposite spins come just in contact when one applies the shift
corresponding to the wavevector $ {\bf q}$ of the FFLO phase.
Therefore this singularity is a general feature of 2D physics and we
expect it to give rise to similar consequences in more realistic and more
complex models describing actual physical systems.
%\vspace{4mm}
* Laboratoire associ\'e au Centre National
de la Recherche Scientifique et aux Universit\'es Paris 6 et Paris 7.
\begin{references}
\bibitem{ff}P. Fulde and R. A. Ferrell, Phys.Rev. {\bf 135}, A550
(1964).
\bibitem{larkov}A. I. Larkin and Y. N. Ovchinnikov, ZhETF {\bf 47},
1136 (1964) [Sov. Phys. JETP {\bf 20}, 762 (1965)].
\bibitem{hbbm}M. Houzet, A. Buzdin, L. Bulaevskii and M. Maley,
Phys. Rev. Lett. {\bf 88}, 227001 (2002).
\bibitem{rcffan}R. Combescot, Europhys. Lett.{\bf 55}, 150 (2001).
\bibitem{bowers}J. A. Bowers and K. Rajagopal, Phys. Rev. D {\bf
66} 065002 (2002) and
references therein.
\bibitem{casal}R. Casalbuoni and G. Nardulli, to be published in Rev.
Mod. Phys.
\bibitem{nam}M. S. Nam , J. A. Symington, J. Singleton, , S. J.
Blundell, A. Ardavan, J.A.A.J. Perenboom, M. Kurmoo and P. Day,
J. Phys. Cond. Mat. {\bf 11}, L477 (1999).
\bibitem{singleton}J. Singleton, J. A. Symington, M. S. Nam, A.
Ardavan, M. Kurmoo and P. Day, J. Phys. Cond. Mat. {\bf 12}, L641
(2000).
\bibitem{tanatar}M. A. Tanatar, T. Ishiguro, H. Tanaka and H.
Kobayashi, cond-mat/0205239.
\bibitem{uji}S. Uji \emph{et al.} Nature {\bf 410}, 908 (2001).
\bibitem{bul}L. N. Bulaevskii, ZhETF {\bf 65}, 1278 (1973) [Sov.
Phys. JETP {\bf 38}, 634 (1974)].
\bibitem{shirai}H. Shimahara and D. Rainer, J. Phys. Soc. Japan, {\bf
66}, 3591 (1997).
\bibitem{manalo}S. Manalo and U. Klein, Phys.Rev. B{\bf 65},
144510 (2002).
\bibitem{klrashi}U. Klein, D. Rainer and H. Shimahara, J. Low.
Temp. Phys. {\bf 118}, 91 (2000).
\bibitem{br}H. Burkhardt and D. Rainer, Ann.Physik {\bf 3}, 181
(1994).
\bibitem{matsuo}S. Matsuo, S. Higashitani, Y. Nagato and K. Nagai,
J. Phys. Soc. Japan, {\bf 67} 280 (1998).
\bibitem{buz1}M. Houzet, Y. Meurdesoif, O. Coste and A. I. Buzdin,
Physica C {\bf 316}, 89 (1999).
\bibitem{cm}R. Combescot and C. Mora, Eur.P.J. B {\bf 28}, 397
(2002).
\bibitem{shima2}H. Shimahara, J. Phys. Soc. Japan, {\bf 67}, 736
(1998).
\bibitem{shima1}H. Shimahara, Phys. Rev. B, {\bf 50}, 12760
(1994).
\bibitem{longpaper}Details will be published elsewhere.
\end{references}
\end{multicols}
\end{document}