Phase diagram of orbital-selective Mott transitions at finite temperatures
Extracted source text ยท arXiv cond-mat/0506151
\documentclass[twocolumn]{jpsj2}
\def\runtitle{Instructions for the Preparation of Manuscript}
\def\runauthor{Online-Journal Subcommittee}
\usepackage{graphicx}% Include figure files
\usepackage{dcolumn}% Align table columns on decimal point
\usepackage{bm}% bold math
\title{Phase diagram of orbital-selective Mott transitions
at finite temperatures}
\author{Kensuke {\sc Inaba},
Akihisa {\sc Koga},
Sei-ichiro {\sc Suga} and
Norio {\sc Kawakami}}
\inst{%
Department of Applied Physics, Osaka University, Suita, Osaka 565-0871 %, Japan
}%
\recdate{\today}% It is always \today, today,
% but any date may be explicitly specified
\abst{
Mott transitions in the two-orbital Hubbard model with different
bandwidths are investigated at finite temperatures.
By means of the self-energy functional approach, we discuss the
stability of the intermediate phase with one orbital localized
and the other itinerant,
which is caused by the orbital-selective Mott transition (OSMT).
It is shown that the OSMT realizes two different coexistence regions
at finite temperatures in accordance with the recent results of
Liebsch.
%% which reflects the fact that the Mott transition for
%%each orbital stabilizes the insulating state independently.
We further find that the particularly interesting behavior
emerges around the special condition $U=U'$ and $J=0$,
which includes a new type of the coexistence region with three
distinct states. By systematically changing the
Hund coupling, we establish the global phase diagram to
elucidate the key role played by the Hund coupling on the
Mott transitions.
}
\kword{Orbital-selective Mott transition, Self-energy functional approach}
%Use showkeys class option if keyword
%display desired
\begin{document}
\maketitle
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% introduction %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%
%\section{Introduction}\label{sec:int}
%%%%%%%%%%%%%%%%%%%%%%
Strongly correlated electron systems with some orbitals have
been investigated extensively.\cite{ImadaRev,TokuraScience}
In particular, substantial progress in theoretical
understanding of Mott transitions of multi-orbital systems
has been made by dynamical mean-field theory (DMFT) \cite{GeorgesRev,KotliarPT} calculations. \cite{Kim90,Kotliar96,Rozenberg97,Bunemann:Gutzwiller,Hasegawa98,Held98,Han98,Momoi98,Klejnberg98,Imai01,Koga:ED_DMFT,Oudovenko02,Ono03,Tomio04,Pruschke04,Sakai04,Liebsch,Koga04,Koga05,Ferrero05,Medici05,Arita05,Knecht05,note,Biermann05}
Among them, the orbital-selective Mott transition (OSMT)\cite{Anisimov02}
in the multi-orbital system with different bandwidths has been
one of the most active topics in this context.
Typical materials are
$\rm Ca_{2-x}Sr_xRuO_4$\cite{Nakatsuji} and
$\rm La_{n+1}Ni_nO_{3n+1}$\cite{LaNiO,Kobayashi96}, where
the OSMT is suggested to be realized by the chemical substitution
and the change in the temperature.
These experimental findings have stimulated theoretical
investigations on this issue.
\cite{Liebsch,Sigrist04,Fang04,Anisimov02,Koga04,Koga05,Ferrero05,Medici05,Arita05,Knecht05,note,Biermann05}
It has been clarified
by extensive works that in the two-orbital system with different
bandwidths at zero temperature, the OSMT occurs in general and merges to
a single Mott transition in a certain restricted parameter region.
\cite{Koga04,Ferrero05,Medici05}
Also,
it has been realized that the Hund coupling is an important key parameter
to understand the correct nature of the OSMT.\cite{KogaPr,note}
In contrast to the detailed analysis at zero temperature,
systematic discussions at finite temperatures
\cite{Koga05,Liebsch,note,Knecht05}
are lacking.
In particular, the stability of the intermediate metallic phase
may be important to discuss the possibility of the OSMT in real
materials. It is thus desirable to investigate the OSMT
in the multi-orbital model at finite temperatures.
Quite recently, Liebsch have studied finite-temperature properties
of the OSMT by DMFT combined with
the exact diagonalization method,\cite{note} where the OSMT
occurs even at finite temperatures.
In this paper, we give the detailed investigation of
the OSMT in the two-orbital Hubbard model at zero and finite temperatures.
We make use of the self-energy functional approach (SFA) proposed by Potthoff,\cite{SFA}
which is efficient to study finite-temperature properties,
to determine the detailed phase diagram.
The present study provides the results complementary to
Liebsch's in a certain parameter regime,\cite{note} and moreover establishes
rich phase diagrams, which include a new type of the
coexistence region with three distinct states.
We further clarify how the magnitude of the Hund coupling controls
the nature of phase diagrams at finite temperatures.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% model %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\section{Model and Method}\label{sec:met}\label{sec:mod}
We consider the two-orbital Hubbard model with different bandwidths,
which is given by the
Hamiltonian, ${\cal H}={\cal H}_0+\sum_i {\cal H}_i^\prime$ with
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{eqnarray}
{\cal H}_0&=&\sum_{<i,j>, \alpha, \sigma}%\sum_{\alpha}\sum_{\sigma}
\left( t_\alpha -\mu\delta_{ij}\right)
c^\dag_{i\alpha\sigma} c_{j\alpha\sigma},\\
%{\cal H}^\prime&=&\sum_i{\cal H}^{\prime}(i)\nonumber\\
%{\cal H}^{\prime}(i)&=&
{\cal H}_i^\prime&=& \!\!\!\!
U \sum_{\alpha}
n_{i \alpha \uparrow} n_{i \alpha \downarrow}
+\sum_{\sigma \sigma^\prime}
(U^\prime - \delta_{\sigma\sigma^\prime}J)
n_{i 1 \sigma} n_{i 2 \sigma^\prime}
\nonumber\\
&-& \!\!\!\!\! J (c^\dag_{i 1 \uparrow}c_{i 1 \downarrow}
c^\dag_{i 2 \downarrow}c_{i 2 \uparrow}
+c^\dag_{i 1 \uparrow}c^\dag_{i 1 \downarrow}
c_{i 2 \uparrow}c_{i 2 \downarrow}+H.c.)
\label{eq:model},
\end{eqnarray}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
where $c^\dag_{i\alpha\sigma}(c_{i\alpha\sigma})$ is
the creation (annihilation) operator of an electron at the $i$th site
with spin $\sigma(=\uparrow,\downarrow)$ and orbital $\alpha(=1,2)$,
and $n_{i\alpha\sigma}=c^\dag_{i\alpha\sigma}c_{i\alpha\sigma}$.
Here, $t_\alpha$ denotes the hopping integral for orbital $\alpha$,
$\mu$ the chemical potential,
$U (U')$ the intra-orbital (inter-orbital) Coulomb interaction,
and $J$ the Hund coupling including the spin-flip and pair-hopping terms.
In the following, we impose the condition $U=U'+2J$,
which results from rotational symmetry of degenerate orbitals.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% method %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%===================================================================
%== outline of SFA ==
%===================================================================
In order to discuss the Mott transitions at zero and finite temperatures,
we use here the self-energy functional approach SFA,\cite{SFA}
%proposed by Potthoff,
which is based on the Luttinger-Ward variational method.\cite{Luttinger}
This approach allows us to deal with
finite-temperature properties of the multi-orbital system efficiently,
\cite{Inaba05}
where standard DMFT with numerical techniques
may encounter some difficulties in a practical computation when
the number of orbitals increases.
%A remarkable point is that the critical behavior
% can be discussed more precise
%than the DMFT analysis,\cite{GeorgesRev}
%when one chooses the same type of the effective impurity model.
%In fact, by comparing the results obtained by the two-site
%DMFT,\cite{Bulla00,Potthoff01}
%the critical point for the Hubbard model with and
%without the orbital degeneracy
%obtained by SFA with the two-site model \cite{SFA}
%is in good agreement with that obtained by
%DMFT with the aid of numerical
%techniques.\cite{Caffarel94,Moeller95,Bulla99}
In SFA, one makes use of the fact that the Luttinger-Ward functional does not depend on
the detail of the Hamiltonian ${\cal H}_0$ as far as the interaction term
${\cal H}^\prime$ is unchanged.\cite{SFA}
This enables us to introduce a proper reference system with
the same interaction term.
One of the simplest models for the reference system is explicitly given
by the following Hamiltonian,
${\cal H}_{\rm ref}=\sum_i{\cal H}_{\rm ref}^{(i)}$,
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{eqnarray}
%{\cal H}_{\rm ref}&=&\sum_i{\cal H}_{\rm ref}^{(i)},\\
%--------------- Hamiltonian of reference system --------------
{\cal H}_{\rm ref}^{(i)}&=&\sum_{\alpha \sigma } \left[
\epsilon^{(i)}_{0\alpha}
% c^\dag_{i\alpha\sigma} c_{i\alpha\sigma}+\sum_{k=1}^{N_b}
% \varepsilon^{(i)}_{k \alpha}
% a^{(i)\dag}_{k\alpha\sigma}a^{(i)}_{k\alpha\sigma}\right]\nonumber\\
c^\dag_{i\alpha\sigma} c_{i\alpha\sigma}+
\epsilon^{(i)}_{\alpha}
a^{(i)\dag}_{\alpha\sigma}a^{(i)}_{\alpha\sigma}\right]\nonumber\\
%%%%%%%%%%%%%%
% &+&\sum_{k=1}^{N_b}\sum_{\alpha \sigma}V^{(i)}_{k \alpha}
% (c^\dag_{i\alpha\sigma}a^{(i)}_{k\alpha\sigma}+H.c.)\nonumber\\
&+&\sum_{\alpha \sigma}V^{(i)}_{\alpha}
(c^\dag_{i\alpha\sigma}a^{(i)}_{\alpha\sigma}+H.c.)
+H_i^\prime,\label{eq:ref_model}
\end{eqnarray}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
where $a^{(i)\dag}_{\alpha\sigma}(a^{(i)}_{\alpha\sigma})$
creates (annihilates) an electron with $\sigma$ spin and $\alpha$ orbital,
which is connected to the $i$th site in the original lattice.
This approximation is regarded as a finite-temperature extension
of the two-site DMFT\cite{Potthoff01} in some respects.
In this paper, we fix the parameters
$\epsilon_{0\alpha}=0, \epsilon_{\alpha}=\mu$ and $\mu=U/2+U'-J/2$
to discuss the zero and finite temperature properties at half filling.
%on the paramagnetic state.
By choosing the parameters $V_{\alpha}$ to minimize the grand potential,
$\partial\Omega/\partial V_\alpha=0$ $(\alpha=1,2)$,
we can find a proper reference Hamiltonian,
which approximately describes the original correlated system.
We note that the hybridization $V_\alpha$ between a given site on the
original lattice and a site in the reference system may be regarded as
the renormalized bandwidth of the $\alpha$-orbital band.
For example, for small $V_\alpha$, heavy quasi-particles are formed
in the $\alpha$ band, and at $V_\alpha =0$ the system
is driven to the Mott insulating state.
By calculating hybridizations $V_1$ and $V_2$,
we thus discuss the stability of the metallic state in
our two-orbital Hubbard model.
In the following, we focus on the system with $W_1=2.0$ and $W_2=4.0$
to discuss Mott transitions,
where $W_\alpha$ is the bandwidth of $\alpha$-orbital
for the semi-circular density of states (DOS),
$\rho_\alpha(x)=4/\pi W_\alpha \sqrt{1-(2x/W_\alpha)^2}$.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\section{Results}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
We begin our discussions with the Mott transitions
at zero temperature for a typical choice of
the ratio of the parameters, $U'=0.5U$ and $J=0.25U$.
Figure \ref{fig:zero:ent} shows the contour plots of
the grand potential and the corresponding
entropy per site $S/L$ at the stationary point, where
$L$ is the number of sites.
When the Coulomb interaction is small $(U=2.3)$,
the stationary point, where the grand potential is minimized,
is located around
$(V_1, V_2)\sim (0.14, 0.55)$.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[htb]
\includegraphics[width=\linewidth]{ent_omg.eps}
\caption{
The entropy per site $S/L$ as a function of $U$ with fixed ratios
$U'=0.5U$ and $J=0.25U$ at $T=0$.
Inset shows the contour plot of the grand potential %per site
for (a)$U=2.3$, (b) 2.8 and (c) 3.6, where crosses denote the minima
of the grand potential.}
\label{fig:zero:ent}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
This implies that the effective bandwidth for each orbital state is
finite, stabilizing the metallic (M) phase.
Increasing the interaction, the stationary point of the system
touches the $V_2$-axis as shown in the inset (b):
the effective bandwidth for the narrower band $V_1$ vanishes while
$V_2$ still remains finite.
Therefore, the OSMT occurs, where
one band is insulating and the other is still metallic.%\cite{KogaOS}
We refer to
this phase as the OSM phase in the
following. It is seen that the residual entropy
in the OSM phase is $S/L = \log 2$, implying the formation of
free localized $s=1/2$ spins.
Further increase of the interaction induces the Mott transition
in the other band to the Mott insulating (MI) phase,
where $V_1=V_2=0$ as shown in the inset (c).
In this case, the localized spins for two bands form the triplet
state due to the Hund coupling, so that the residual entropy $S/L=\log 3$
in the region $U>3.4$. Note that the residual entropy at zero temperature
is an artifact of our simplified SFA approach. This pathological behavior can be
somehow improved, e.g. by taking into account spin fluctuations
more accurately.
At zero temperature, we can find no other local minima
in the grand potential when the interaction is varied.
Therefore, we clearly observe double second-order Mott
transitions in the multi-orbital system, which indeed confirms
the claim of Koga {\it et al}.\cite{Koga04} for
the OSM transition obtained
by means of DMFT combined with the exact diagonalization.
We now move to the phase diagram at finite temperatures.
Since each transition at zero
temperature is similar to that for the single-orbital Hubbard model,
\cite{GeorgesRev}
the first-order transition may occur at finite temperatures
around each critical point, as shown in ref. \citen{note}.
In fact, we find that when the system belongs to the metallic phase
close to the critical point, two stationary points representing the M and
OSM states appear simultaneously once the system is at finite temperatures
($T=0.002$), as shown in Fig. \ref{fig:local}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[ht]
\includegraphics[width=.9\linewidth]{ent_hys.eps}
\caption{(a) The contour plot of the grand potential $\Omega(V_1,V_2)$
at $T=0.002,U=2.3$ and $J=0.25U$. Two minima coexist, which correspond to
the M and OSM states.
(b) The entropy vs the interaction $U$ at $T=0.002, J=0.25U$.
}
\label{fig:local}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
The double-well structure causes
the first-order Mott transition with hysteresis at finite temperatures
(Fig. \ref{fig:local} (b)). At $T=0.002$, as $U$ increases,
the stationary point for the metallic
state disappears around $U_{c2}\sim 2.36$,
where the Mott transition occurs to the OSM phase.
On the other hand, as $U$ decreases, the OSM phase realized is stable
down to
%below
$U_{c1}\sim 2.24$.
The first-order transition point $U_c\sim 2.33$ for $T=0.002$ is determined
by the crossing point of the two minima in the grand potential.
The obtained phase diagram is shown in Fig. \ref{fig:phase_J25}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[ht]
\includegraphics[width=\linewidth]{phase_J25.eps}
\caption{$U-T$ phase diagram at $J=0.25U$.
There are two coexistence phases, which are featured by the
triangular-shaped regions: ({\it left})the metallic phase and
orbital-selective Mott phase coexist and ({\it right}) orbital-selective
Mott phase and insulating phase coexist.}
\label{fig:phase_J25}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
It is found that the two distinct coexistence regions
appear around $U\sim 2.4$ and $U\sim 3.3$.
The phase boundaries $U_{c1}$, $U_{c2}$ and $U_c$ merge
at the critical temperature $T_c$ for each transition.
Similar phase diagram
was recently obtained by Liebsch by means of DMFT with
the exact diagonalization method, who gave better estimates
of the critical temperatures.\cite{note} Nevertheless, our SFA treatment
elucidates further interesting
properties such as the crossover behavior among the competing
M, OSM and MI phases. To make this point more explicit,
we calculate thermodynamic quantities
such as the entropy and the specific heat, which are
displayed in Fig. \ref{fig:hcap}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[t]
\includegraphics[width=\linewidth]{hcap.eps}
\caption{(a) Entropy $S/L$ and (b) Specific heat $C/L$
as a function of $U$ in the crossover region, $J=0.25U$.
}\label{fig:hcap}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
We can see a double-step structure
similar to that found at zero temperature
in the curve of the entropy. Such anomalies
are more clearly seen in the specific heat. It is quite impressive
that the crossover behavior among three phases are clearly seen
even at higher temperatures. We can thus say that
the OSM phase is rather well defined even at higher temperatures
above the critical temperatures.
%Here, we wish to make a comment on the specific heat:
%the peak structure around $U\sim3.1$ is somehow smeared in
%comparison with that around $U\sim2.2$. This suggests that the crossover
%is slightly different from each other.
%Around $U\sim2.2$, one of the orbitals is almost localized
%whose crossover is essentially the same as that for the single band model.
%On the other hand, in the crossover around $U\sim3.1$,
%the Hund coupling plays an important role and has
%a tendency to make triplet state, which
%obscures the sharp crossover behavior around $U\sim3.1$.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% effect of Hund coupling %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
We have so far discussed the OSMT
for the system with rather large Hund couplings.
According to the zero-temperature analysis,\cite{Koga04,Ferrero05,Medici05}
it is known that around
the special condition $U=U'$ and $J=0$ at $W_1/W_2=0.5$,
a non-trivial single Mott transition occurs, whose origin is
attributed to enhanced orbital fluctuations.\cite{Koga05}
Therefore, it is interesting to observe what happens
at finite temperatures, when the system approaches the
condition $U=U'$ and $J=0$.
By performing similar calculations, we end up with the phase diagram for
the system with $U'=0.94U, J=0.03U$ (Fig. \ref{fig:phase_J3}).
Since the Hund coupling is very small in this case,
two critical points for the OSMT
at zero temperature are close to each other.
One of the most remarkable findings here is that
three competing states (M, OSM and MI states)
coexist in a certain region at finite temperatures,
where three distinct local minima appear
in the grand potential, as shown in the inset.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[ht]
\includegraphics[width=\linewidth]{phase_J3.eps}
\caption{$U-T$ phase diagram at $J=0.03U$: we have used the
same symbols as in Fig. \ref{fig:phase_J25}.
Note that the two coexistence regions are not separated here in
contrast to Fig. \ref{fig:phase_J25}.
Three distinct states (M,OSM,and MI states) coexist
in the region which is surrounded by filled and open circles.
Inset shows the contour plot of the ground potential in the
coexistence region for $U=4.45,T=0.004$, where there are three
local minima.
}\label{fig:phase_J3}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%Furthermore, there is another coexistence region with
% M and MI states, which are not adiabatically connected
%to each other at zero temperature.
In these parameters,
the coexistence regions are quite sensitive to how large
the Hund coupling is.
%In this case,
%the first coexisting region,
%which is made from lines $U_{c1}^{(1)}$ and $U_{c2}^{(1)}$,
%is involved by the second coexisting region is made from
%lines $U_{c1}^{(2)}$ and $U_{c2}^{(2)}$.
%Further pronounced feature affected by the Hund coupling $J$
%is the second coexisting region grows up
%while the first coexisting region remains same scale.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
To observe how the Hund coupling controls finite-temperature
properties, we show the global phase
diagrams in Fig. \ref{fig:phase_all} by systematically changing $J$.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[t]
\includegraphics[width=.9\linewidth]{phase_all.eps}
\caption{Overall picture for the coexistence regions with
systematic change of $J$: two types of the regions
are shown for $J=0.25U, J=0.1U$, $J=0.03U$ and $J=0$ from
left to right.
At $J=0$, the transition from the metal to OSM disappears.
}\label{fig:phase_all}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
As the Hund coupling decreases,
both the critical points $U_{c2}^{(1)}$ and $U_{c2}^{(2)}$
increase monotonically, stabilizing the metallic state
up to fairly large Coulomb interactions, as already
mentioned for the zero-temperature case.\cite{Koga04} A remarkable feature
in Fig. \ref{fig:phase_all} is that the completely different
behavior emerges in the critical temperatures $T_c^{(1)}$ and $T_c^{(2)}$.
It is seen that $T_c^{(1)}$ is almost
unchanged, whereas $T_c^{(2)}$ strongly depends on the Hund coupling.
This may be explained as follows. Around the first
boundary between the M and OSM phases, the
nature of the transition is essentially the same as
the single-orbital case, as discussed above. Therefore,
the Hund coupling little affects the critical temperature
$T_c^{(1)}$. On the other hand, around
the second phase transition between the OSM and
MI phases, orbital fluctuations should play a vital role.
The presence of the Hund coupling somewhat suppresses
orbital fluctuations due to the formation of the triplet state.\cite{Koga:ED_DMFT}
Therefore, as $J$ decreases, enhanced orbital fluctuations raise
the critical temperature $T_{c}^{(2)}$, as seen in Fig.
\ref{fig:phase_all}.
Note that for very small Hund couplings
$(J\lesssim 0.03U)$ the OSM phase disappears,\cite{Medici05} and thus
the finite as well as zero-temperature properties
can be approximately described by the multi-orbital model
$(U\neq U')$ with the same bandwidths.
\cite{Koga:ED_DMFT,Inaba05,Ono03,Pruschke04}
In summary, we have investigated Mott transitions in the
two-orbital Hubbard model with different bandwidths at finite
temperatures. By means of the self-energy functional approach,
we have confirmed that two distinct coexistence regions appear
in the phase diagram in accordance with the recent results obtained
by DMFT. A notable new finding is that the OSM phase is rather well
defined even at temperatures higher than the two critical
temperatures. Further remarkable features have been elucidated in the
phase diagram around special conditions with small Hund couplings
$J \simeq 0$ ($U \simeq U'$), where the coexistence region
with three competing phases emerges. Orbital
fluctuations are enhanced there,
and therefore the system gets very sensitive to
small perturbations, giving rise to the rich
phase diagram at low temperatures.
We have also clarified the role of the Hund coupling
on global features in the phase diagram: the Hund coupling has
little (sizable) effect on the transition between the M and OSM
(OSM and MI) phases, which is again attributed to the role played by
enhanced orbital fluctuations.
In this paper we have exploited the simplified version of SFA to
elucidate fundamental properties at finite temperatures. Since
instabilities to possible ordered phases have not been considered here,
SFA is to be generalized to incorporate such ordered phases, which
should be done in the future work.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% acknowledgments %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%\begin{acknowledgments}
We would like to thank M. Sigrist, T. M. Rice
and A. Liebsch for valuable discussions.
Numerical computations were carried out at the Supercomputer Center,
the Institute for Solid State Physics, University of Tokyo.
This work was supported by a Grant-in-Aid for Scientific Research from
the Ministry of Education, Culture, Sports, Science, and Technology, Japan.
%%\end{acknowledgments}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% REFERENCES
%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\bibliography{main}
%
\begin{thebibliography}{10}
\expandafter\ifx\csname url\endcsname\relax
\def\url#1{\texttt{#1}}\fi
\expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
\bibitem{ImadaRev}
M.~Imada, A.~Fujimori and Y.~Tokura: Rev. Mod. Phys. {\bf 70} (1998) 1039.
\bibitem{TokuraScience}
Y.~Tokura and N.~Nagaosa: Science {\bf 288} (2000) 462.
\bibitem{GeorgesRev}
A.~Georges, G.~Kotliar, W.~Krauth and M.~J. Rozenberg: Rev. Mod. Phys. {\bf 68}
(1996) 13.
\bibitem{KotliarPT}
G.~Kotliar and D.~Vollhardt: Phys. Today {\bf 57} (2004) 53.
\bibitem{Kim90}
C.-I. Kim, Y.~Kuramoto and T.~Kasuya: J. Phys. Soc. Jpn. {\bf 59} (1990) 2414.
\bibitem{Kotliar96}
G.~Kotliar and H.~Kajueter: Phys. Rev. B {\bf 54} (1996) 14221.
\bibitem{Rozenberg97}
M.~J. Rozenberg: Phys. Rev. B {\bf 55} (1997) R4855.
\bibitem{Bunemann:Gutzwiller}
J. B{\"{u}}nemann and W. Weber, Phys. Rev. B {\bf 55}, 4011 (1997); J.
B{\"{u}}nemann and W. Weber and F. Gebhard, {\it ibid} {\bf 57}, 6896 (1998).
\bibitem{Hasegawa98}
H.~Hasegawa: J. Phys. Soc. Jpn. {\bf 56} (1997) 1196.
\bibitem{Held98}
K.~Held and D.~Vollhardt: Eur. Phys. J. B {\bf 5} (1998) 473.
\bibitem{Han98}
J.~E. Han, M.~Jarrel and D.~L. Cox: Phys. Rev. B {\bf 58} (1998) R4199.
\bibitem{Momoi98}
T.~Momoi and K.~Kubo: Phys. Rev. B {\bf 58} (1998) R567.
\bibitem{Klejnberg98}
A.~Klejnberg and J.~Spalek: Phys. Rev. B {\bf 57} (1998) 12401.
\bibitem{Imai01}
Y.~Imai and N.~Kawakami: J. Phys. Soc. Jpn {\bf 70} (2001) 2365.
\bibitem{Koga:ED_DMFT}
A. Koga and Y. Imai and N. Kawakami: Phys. Rev. B {\bf 66} (2002) 165107; J.
Phys. Soc. Jpn. {\bf 72} (2003) 1306.
\bibitem{Oudovenko02}
V.~S. Oudovenko and G.~Kotliar: Phys. Rev. B {\bf 65} (2002) 075102.
\bibitem{Ono03}
Y.~Ono, M.~Potthoff and R.~Bulla: Phys. Rev. B {\bf 67} (2003) 035119.
\bibitem{Tomio04}
Y.~Tomio and T.~Ogawa: ond-mat/0407314.
\bibitem{Pruschke04}
T.~Pruschke and R.~Bulla: Eur. Phys. J. B {\bf 44} (2005) 217.%cond-mat/0411186.
\bibitem{Sakai04}
S.~Sakai, R.~Arita and H.~Aoki: cond-mat/0405503.
\bibitem{Liebsch}
A. Liebsch, Phys. Rev. Lett. {\bf 91}, 226401 (2003); Europhys. Lett. {\bf 63},
97 (2003); Phys. Rev. B {\bf 70}, 249904 (2004).
\bibitem{Koga04}
A.~Koga, N.~Kawakami, T.~M. Rice and M.~Sigrist: Phys. Rev. Lett. {\bf 92}
(2004) 216402.
\bibitem{Koga05}
A.~Koga, N.~Kawakami, T.~M. Rice and M.~Sigrist: cond-mat/0503651.
\bibitem{Ferrero05}
M.~Ferrero, F.~Becca, M.~Fabrizio and M.~Capone: cond-mat/0503759.
\bibitem{Medici05}
L.~de' Medici, A.~Georges and S.~Biermann: cond-mat/0503764.
\bibitem{Arita05}
R.~Arita and K.~Held: cond-mat/0504040.
\bibitem{Knecht05}
C.~Knecht, N.~Bl{\"u}mer and P.~G.~J. van Dongen: condmat/0505106.
\bibitem{note}
After the completion of the present work, we became aware of the recent
preprint; A. Liebsch: cond-mat/0505393.
\bibitem{Biermann05}
S.~Biermann, L.~de' Medici and A.~Georges: cond-mat/0505737.
\bibitem{Anisimov02}
V.~I. Anisimov, I.~A. Nekrasov, D.~E. Kondakov, T.~M. Rice and M.~Sigrist: Eur.
Phys. J. B {\bf 25} (2002) 191.
\bibitem{Nakatsuji}
S. Nakatsuji {\it et al}.: Phys. Rev. Lett. {\bf 90} (2003) 137202; S.
Nakatsuji and Y. Maeno: Phys. Rev. Lett. {\bf 84} (2000) 2666.
\bibitem{LaNiO}
K. Sreedhar: {\it et al.}, J. Solid State Comm. {\bf 110} (1994) 208; Z. Zhang,
{\it et al.}: J. Solid State Comm. {\bf 108} (1994) 402; {\bf 117} (1995)
236.
\bibitem{Kobayashi96}
Y.~Kobayashi, S.~Taniguchi, M.~Kasai, M.~Sato, T.~Nishioka and M.~Kontani: J.
Phys. Soc. Jpn {\bf 65} (1996) 3978.
\bibitem{Sigrist04}
M.~Sigrist and M.~Troyer: Eur. Phys. J. B {\bf 39} (2004) 207.
\bibitem{Fang04}
Z.~Fang, N.~Nagaosa and K.~Terakura: Phys. Rev. B {\bf 69} (2004) 45116.
\bibitem{KogaPr}
A.~Koga, N.~Kawakami, T.~M. Rice and M.~Sigrist: cond-mat/0406457.
\bibitem{SFA}
M. Potthoff: Eur. Phys. J. B {\bf 32} (2003) 429; M. Potthoff: {\it ibid} {\bf
36} (2003) 335; K.~Pozg{\v{a}j\v{c}i\'{c}}: cond-mat/0407172.
\bibitem{Luttinger}
J.~M. Luttinger: Phys. Rev. {\bf 118} (1960) 1417.
\bibitem{Inaba05}
K.~Inaba, A.~Koga, S.~Suga and N.~Kawakami: cond-mat/0506150.
\bibitem{Potthoff01}
M.~Potthoff: Phys. Rev. B {\bf 64} (2001) 165114.
\end{thebibliography}
\end{document}