1: %\documentclass[12pt,a4j]{sci}%% for Science
2: \documentclass[prl,twocolumn,floatfix,showpacs,superscriptaddress]{revtex4}
3: %\documentclass[prl,preprint,floatfix,showpacs,superscriptaddress]{revtex4}
4: %\documentclass[aps,prl,twocolumn]{revtex4}
5: \usepackage{amsmath}
6: \usepackage[dvips]{graphicx}% Include figure files
7:
8: %\textheight=22.0cm
9: %\textwidth=16.0cm
10: %\parindent=1cm
11: %\renewcommand{\baselinestretch}{1.5}
12:
13: %-------------------------------------------------------------------------
14: % Define a new command to open an equation mode WITH numeration in the paper
15: % and label
16: % e.g. \eq{<equation>}{<label >}
17: %-------------------------------------------------------------------------
18: \newcommand{\eq}[2]
19: {\begin{equation}
20: #1
21: \label{#2}
22: \end{equation}}
23:
24: % The same with no label in the paper!!
25: \newcommand{\eqnn}[2]
26: {\begin{equation*}
27: #1
28: \label{#2}
29: \end{equation*}}
30:
31:
32:
33: \begin{document}
34:
35: \title{Metal-insulator transitions in the Periodic Anderson Model}
36: \author{G. Sordi}
37: \affiliation{Laboratoire de Physique des Solides, CNRS-UMR8502, Universit\'e de Paris-Sud,
38: Orsay 91405, France.}
39: \author{A. Amaricci}
40: \affiliation{Laboratoire de Physique des Solides, CNRS-UMR8502, Universit\'e de Paris-Sud,
41: Orsay 91405, France.}
42: \author{M.J. Rozenberg}
43: \affiliation{Laboratoire de Physique des Solides, CNRS-UMR8502, Universit\'e de Paris-Sud,
44: Orsay 91405, France.}
45: \affiliation{Departamento de F\'{\i}sica, FCEN, Universidad de Buenos Aires,
46: Ciudad Universitaria Pab.I, Buenos Aires (1428), Argentina.}
47:
48: \date{today}
49: \begin{abstract}
50: We solve the Periodic Anderson model in the Mott-Hubbard regime,
51: using Dynamical Mean Field Theory.
52: Upon electron doping of the Mott insulator, a metal-insulator
53: transition occurs which is qualitatively similar to
54: that of the single band Hubbard model,
55: namely with a divergent effective mass and a first
56: order character at finite temperatures.
57: Surprisingly, upon hole doping, the metal-insulator transition
58: is not first order and does not show a divergent mass.
59: Thus, the transition scenario of the single band Hubbard model is not generic
60: for the Periodic Anderson model, even in the Mott-Hubbard regime.
61: \end{abstract}
62:
63:
64: \pacs{71.30.+h,71.10.Fd,71.27.+a}
65:
66: \maketitle
67:
68: The metal-insulator transition in strongly correlated materials remains
69: a central problem of modern condensed matter physics \cite{mott,ift}.
70: Great progress in its understanding was made possible by the development
71: of new theoretical approaches such as the
72: Dynamical Mean Field Theory \cite{rmp},
73: which is a method that becomes exact in the limit of large lattice connectivity
74: \cite{mv}.
75: The mean field equations can usually be tackled with a variety of numerical
76: approaches which allow to obtain reliable solutions and insights.
77: In this context, the Hubbard model, which is probably the simplest model that
78: captures a correlation driven metal-insulator transition (MIT), called Mott-Hubbard
79: transition,
80: has received most of the attention in the past 15 years.
81: As a result of intense investigation, our understanding of the metal-insulator
82: transition in that model is now profound.
83: The studies have unveiled a scenario where, at low temperatures and moderate
84: interaction, the half-filled Mott insulator may be driven to a correlated
85: metallic state through a first order transition\cite{ucs}.
86: The transition can occur as a function of correlation strength, temperature
87: or doping. The first order line ends at finite temperature in a critical point
88: and the critical region can be described by a Ginzburg-Landau theory \cite{prls}.
89: This theoretical prediction was experimentally verified in
90: experiments on $V_2O_3$ \cite{limelette}.
91: The Hubbard model is often considered as a minimal model
92: for the study of rather complicated compounds such as transition metal oxides
93: and heavy fermion systems.
94: This is supported by the implicit assumption that the Hubbard model is expected
95: to be the effective low energy Hamiltonian for a wider class of more realistic
96: multiband models for strongly correlated electron systems.
97:
98: On the other hand, a more realistic model
99: which is
100: also widely used in theoretical investigations
101: of strongly interacting systems, though still schematic,
102: is the Periodic
103: Anderson model (PAM).
104: In the context of correlated electron systems, this model permits to
105: describe explicitly both, the localized orbitals, such as the $d$ in
106: transition metal oxides or the $f$ in heavy fermion systems, and their
107: hybridization to an itinerant electron band
108: (such as that of $p$ orbitals of oxygen in transition metal oxides).
109: In fact, the PAM allows to investigate the various regimes where Mott insulating
110: states occur, as characterized by the Zaanen-Sawatzky-Allen (ZSA) scheme \cite{zsa}.
111: They are classified as either Mott-Hubbard insulators
112: or charge transfer insulators. The first
113: apply to the early transition metal oxides such as titanates and vanadates, while
114: the second is relevant for cuprates, such as the high $T_c$ superconductors, and
115: manganites, which show colossal magnetoresistance \cite{ift}.
116: In theoretical studies, however, it is often assumed that both
117: Mott-Hubbard and charge transfer systems may be described at
118: low energies by a simpler one band Hubbard model Hamiltonian.
119:
120: In the present work we shall test the
121: putative validity of the Hubbard model as the
122: effective low energy Hamiltonian of the more realistic Periodic Anderson model.
123: We shall do this
124: within a well defined mathematical framework, namely, the
125: Dynamical Mean Field Theory (DMFT),
126: that allows us to obtain essentially exact numerical solutions of the models
127: (in the statistical Monte Carlo sense).
128: In particular we shall concentrate on the nature of the (paramagnetic)
129: metal-insulator transitions
130: that occur in the Periodic Anderson model with parameters that set it
131: in the Mott-Hubbard regime, and discuss it with respect to the corresponding
132: scenario that is realized
133: in the one band Hubbard model case.
134: In addition, our results should also be valuable for the interpretation of
135: experimental spectroscopies of strongly correlated transition metal oxides,
136: that experienced fantastic improvements in the last decade. In fact, the
137: analysis of experimental data of systems which have a mixed orbital
138: character is not always simple when strong correlations are present.
139: Finally, our work addresses a very relevant issue in regard of the intense effort
140: that is currently dedicated to the implementation of {\it ab initio} methods
141: for strongly correlated materials \cite{phystoday} which makes heavy use of the
142: DMFT methodology \cite{gabyRMP}.
143:
144: Among our main results we find that in the case of the electron
145: doped driven MIT, the scenario is indeed similar to the one realized in the
146: Hubbard model, however, the hole doping scenario
147: is qualitatively different. In this case, the correlated metal has
148: a resonance peak at the Fermi energy flanked by a Hubbard band, but,
149: unlike the Hubbard model scenario, it is not related to the formation of a
150: Kondo like resonance and its mass does not diverge at the transition.
151: Moreover, and also in contrast to the Hubbard model case, our results indicate
152: that this metal-insulator transition is of second order as no signs of
153: coexistent solutions were observed.
154: We shall argue that while the metallic state in the former case is a renormalized
155: ``Brinkman-Rice'' Fermi liquid \cite{br}, the latter can be interpreted
156: as liquid of ``Zhang-Rice singlets'' \cite{zr}.
157:
158: The Periodic Anderson model Hamiltonian reads,
159: %
160: \eqnn{
161: \begin{split}
162: H=&-t\sum_{<ij>\sigma} (p^+_{i\sigma}p_{j\sigma} + p^+_{j\sigma}p_{i\sigma}) +
163: \left(\epsilon_p - \mu \right) \sum_{i\sigma} p^+_{i\sigma}p_{i\sigma} \\
164: &+(\epsilon_d-\mu)\sum_{i\sigma} d^+_{i\sigma}d_{i\sigma}
165: +t_{pd} \sum_{i\sigma}\left( d^+_{i\sigma}p_{i\sigma} +p^+_{i\sigma}d_{i\sigma} \right) \\
166: &+U\sum_i \left(n_{di\uparrow}-\tfrac{1}{2}\right)\left(n_{di\downarrow}-\tfrac{1}{2}\right)
167: \end{split}}{PAMHam}
168: %
169: \noindent
170: where the $d_\sigma$ and $d^+_\sigma$ operators destroy and create electrons
171: at non-dispersive $d-$orbitals with energy $\epsilon_d$,
172: $p_\sigma$ and $p^+_\sigma$ destroy and create electrons at $p-$orbitals
173: with energy $\epsilon_p$ which form a band with hopping parameter $t$.
174: The $p-$ and $d-$orbitals are hybridized with an amplitude $t_{pd}$,
175: and the electron correlations are introduced by the Coulomb interaction
176: $U$ on the $d-$sites. It is customary to define the charge transfer
177: energy $\Delta = \epsilon_d - \epsilon_p$, and $\mu$ is the chemical
178: potential. As described in the ZSA scheme this model
179: predicts correlated insulating states in two very different regimes:
180: at $\Delta << U$ the charge transfer insulator, and at $U \lesssim \Delta$
181: the Mott-Hubbard insulator. The latter is relevant for the early transition
182: metal oxides and will be the focus of the present work.
183:
184: %which becomes exact in the case of a large coordination lattice.
185: To solve the PAM using DMFT, for simplicity
186: we adopt a Bethe lattice that corresponds to a semicircular
187: density of states (DOS) for the $p-$electron band. Setting
188: the hopping of the $p-$electrons to $t=1/2$, their
189: half-bandwidth is equal to one, and fixes the units of the
190: model.
191: To set the system in the Mott-Hubbard regime, we adopt
192: $\epsilon_d =0$ and $\epsilon_p$ negative, so that the $p-$band
193: lies well below the Fermi surface
194: and is almost full, while the occupation of the local $d-$sites
195: will be close to one. The parameter $t_{pd}$ controls the hybridization
196: between the orbitals at each lattice site, and permits the delocalization
197: of the $d-$electrons. In fact, a finite $t_{pd}$ turns the ``flat'' band
198: of $d-$orbitals into a conduction band with mainly $d$ character and
199: bandwidth of the order of $t_{pd}^2/\Delta$.
200: Now, for a moderate value of the repulsion $U > t_{pd}^2/\Delta$
201: and an occupation of the $d-$site $n_d$ close to one, one expects
202: the conduction band to open a correlation gap and the system
203: becomes a Mott-Hubbard insulator.
204:
205: The DMFT equations are most easily derived using the cavity
206: method \cite{gks,rmp}, and one obtains the local effective action
207: for the $d-$electrons:
208: \eq{
209: \begin{split}
210: {\rm S_{\rm eff}}=&
211: %
212: -\int_0^{\beta}d\tau \int_0^\beta d\tau' \sum_{\sigma} d^+_{\sigma}(\tau)
213: {\cal G}^{-1}_0(\tau - \tau')d_{\sigma}(\tau') \\
214: %
215: &+U\int_0^{\beta} d\tau \left[n_{d\uparrow}(\tau)-\tfrac{1}{2}\right]
216: \left[n_{d\downarrow}(\tau) -\tfrac{1}{2}\right]
217: %+{\rm const.}
218: \end{split}
219: }{Seff}
220: %
221: where $d_{\sigma}$ and $d^+_{\sigma}$ correspond to a given (any) site of the lattice.
222: This equation defines the so called associated impurity problem of the model, that is
223: subject to a self consistent constraint that reads,
224: %
225: \eqnn{
226: {\cal G}_0^{-1}(i\omega_n)=
227: i\omega_n+\mu-\epsilon_d-\frac{{{t^2_{pd}}}}{i\omega_n+\mu-\epsilon_p-t^2G_{pp}}\ .}
228: {self}
229: %
230: The solution of the quantum impurity problem (\ref{Seff}) gives the local
231: $d-$electron Green's function $G_{dd}$ and defines a self-energy
232: $\Sigma$=${\cal G}^{-1}_0 - G^{-1}_{dd}$.
233: The local Green's function of the $p-$electrons $G_{pp}$ is
234: obtained in terms of $\Sigma$ and the non-interacting
235: semicircular DOS $\rho_0$ as:
236: %
237: $$
238: G_{pp}=\int{ d\epsilon \frac{\rho_0(\epsilon)}{i\omega_n +\mu -\epsilon_p -
239: \frac{t^2_{pd}}{i\omega_n+\mu-\epsilon_d -\Sigma(i\omega_n)} - \epsilon}}\ .
240: $$
241: %
242: \noindent
243:
244: We solve for these equations using two powerful, and in principle
245: exact, numerical methods: Quantum Monte Carlo (QMC)
246: % \cite{hf}
247: and Exact Diagonalization (ED) \cite{rmp}. A similar methodology
248: was used in the study of a related two-band Hubbard model
249: \cite{mutou-ono}.
250: %\cite{ed}.
251: The QMC is a finite temperature method
252: and is exact in the statistical sense, while
253: the ED is at $T=0$ and relies on diagonalization of finite clusters and
254: extrapolations to account for the
255: systematic finite size effects \cite{rmp}.
256:
257:
258:
259: \begin{figure}%[!ht]
260: \centering
261: \includegraphics[width=7cm,clip]{fig1.eps}
262: \caption{
263: Density of states for the $p-$ and $d-$electrons (dashed and solid line)
264: for $\Delta=1$, $t_{pd}=0.9$.
265: Upper panel shows the $U=0$ case with $\mu=0.529$ that gives a total occupation
266: of three particles below the Fermi energy.
267: %The localized $d-$states are broaden into a partially filled narrow band due
268: %to the finite hybridization with
269: %lower $p-$electron band.
270: The lower panel shows analytically continued QMC data for the Mott insulator
271: state with $U=2$, $T=1/64$ and $\mu=1.079$ that
272: gives $n_{tot}=3$.
273: The double arrow head line indicates the large Mott gap $\Delta_{M}$.
274: The inset shows $\Delta_M(U)$ for different positions of the
275: of the $p-$electron band $\epsilon_p=-6,-3,-2,-1$
276: (top to bottom). The results are obtained with ED of finite clusters of $N_s$ sites
277: and the data shown is for the limit $N_s \rightarrow\infty$.
278: }
279: \label{fig1}
280: \end{figure}
281: %
282: We shall begin our discussion with the behavior of the
283: density of states
284: of the model in different regimes.
285: For $U=0$ and at $t_{pd}=0$ the system is an insulator since
286: neither the $d-$orbitals at the Fermi energy can conduct (because
287: they are localized), nor the $p-$band can conduct (because it is
288: full and well beneath the Fermi surface).
289: At a finite hybridization $t_{pd}$, however, the
290: system becomes metallic, as the $p$ and $d$ orbitals form
291: a partially filled band at the Fermi energy with mixed $p$ and $d$ character.
292: In Fig.\ref{fig1} we show the comparison of the $p-$ and $d-$electron DOS.
293: The one carrying most of the
294: spectral intensity at low frequencies is the $d-$electron DOS $\rho_d(\omega)$,
295: since the bare atomic energy of the $d-$orbitals is at
296: the Fermi energy.
297: The lower panel of the figure shows the dramatic effect of
298: correlations; when the interaction $U$ is increased,
299: a rather large gap opens in the DOS at the Fermi energy,
300: driving the system to a Mott insulator state.
301: The gap is of order $U$ and results from the
302: high energetic cost of double occupation of the $d$ site.
303: An interesting effect is that the size of the Mott gap $\Delta_M$
304: may be substantially renormalized. In the inset of Fig.\ref{fig1}
305: we show the variation of $\Delta_M(U)$ upon increasing the distance
306: of the $p-$band with respect to the $d-$electron energy.
307: Notice that the gap $\Delta_M(U)$ is always smaller than the bare $U$,
308: and become equal only asymptotically when $\epsilon_p \to -\infty$.
309: This renormalization effect is of relevance to the difficult problem of the
310: determination of the effective value of $U$ in realistic
311: {\it ab initio} calculations using DMFT \cite{phystoday}.
312:
313: \begin{figure}%[!ht]
314: \centering
315: \includegraphics[width=7cm,clip]{fig2.eps}
316: \caption{$n_d$ (solid line), $n_p$ (dashed) and $n_{tot}$
317: (dotted) as a function of $\mu$, for $U=2$.
318: The data are from QMC at $\Delta=1$, $t_{pd}=0.9$ and $T=1/64$.
319: The plateaux at $n_{tot}$=2 and 4 are band insulator (BI) states.
320: The one at $n_{tot}$=3 is the Mott insulator (MI).
321: The inset shows the phase diagram in the $U$-$\mu$ plane.
322: %for $\Delta=1$ and $t_{pd}=0.9$. The region MI is the Mott insualting phase
323: %at $n_{tot}=3$.
324: The boundary lines are for $T$=1/20 (dotted), $T$=1/64 (thick solid)
325: and $T=0$ (thin solid).
326: The dashed thick line segment at $T=1/64$ denotes the region of the MIT boundary
327: where the QMC data show coexistence of a metal and an insulating solution.
328: The $T$=0 data are from extrapolated ED calculations. This method is not
329: suited for the study of coexistence of solutions at $T=0$.
330: The dash-dot line denotes the transition from a metal (M) to the band insulator
331: (BI) at $n_{tot}=4$.
332: }
333: \label{fig2}
334: \end{figure}
335:
336: The Mott insulator can be destabilized by either
337: particle or hole doping.
338: Therefore the system has two doping driven
339: metal-insulator transitions. In the one band Hubbard model,
340: the two transitions have the same character, however, as we shall
341: see, this is not the case in the present model.
342: In Fig.\ref{fig2} we
343: show the occupation $n_d$ and $n_p$ of the $d$ and $p$ sites
344: for $U=2$.
345: The plateaux that appear around $0.7\lesssim \mu \lesssim 1.2$
346: indicate the onset of the
347: incompressible Mott insulating state when correlations are strong.
348: While the Mott insulator is associated with the energy
349: cost of doubly occupying the local $d-$orbital, it is interesting
350: to notice that the Mott plateau does not occur
351: exactly at $n_d=1$ but at a higher value, which depends on the hybridization.
352: The Mott state is in fact found when the total number of
353: particles per unit cell is exactly equal to three.
354: Thus, the object that becomes localized due to the strong correlations
355: is not simply a $d-$electron, but a composite object with mixed $d$ and
356: $p$ character.
357: The inset of the figure shows the phase diagram in the $U$-$\mu$ plane,
358: that maps the region of the Mott insulator phase and the transition
359: lines to correlated metallic states.
360:
361: \begin{figure}%[!ht]
362: \centering
363: \includegraphics[width=8cm,clip]{fig3.eps}
364: \caption{Density of states of $p-$ and $d-$electrons (dashed and solid line)
365: for $U=2$, $\Delta=1$, $t_{pd}=0.9$ and $T=1/64$, as obtained from QMC.
366: {\it Upper panel}: $\mu=0.554$, which corresponds to tiny hole doping.
367: {\it Lower panel}: $\mu=1.234$, which corresponds to tiny particle doping.}
368: \label{fig3}
369: \end{figure}
370:
371: In Fig.\ref{fig3} we show the DOS for the $p$ and $d$ electrons in the
372: metallic states that are obtained by either particle or hole doping of
373: the Mott insulator. In both cases one finds that the
374: DOS clearly show the emergence of a correlated small
375: quasiparticle peak at the Fermi energy.
376: The occurrence of a narrow quasiparticle peak at the Fermi energy that
377: is flanked by
378: large Hubbard bands which are separated by an energy of order $U$
379: is a hallmark
380: result of the solution of the Hubbard model within DMFT \cite{rmp}.
381: Thus, one may be led to conclude that the MIT in the PAM shares the same
382: qualitative features.
383: Rather surprisingly, this expectation is only fully confirmed for the case
384: of particle doping, but the MIT scenario in the hole doped case is
385: strikingly different.
386: Upon particle doping of the Mott insulating state we have confirmed
387: that there is a small region of parameters at the MIT boundary
388: where two coexistent solutions,
389: one metallic and one insulating, are found.
390: In addition, the numerical solutions show critical
391: slowing down of the convergence speed of the
392: self-consistency close to the transition.
393: These two features were also observed in the previous studies of
394: the finite $T$ first order MIT in the Hubbard model \cite{prls}.
395: In contrast, in the hole doped case ($n_{tot} < 3$) we found no trace of
396: coexistent solutions down to $T=1/128$.
397: The solution seems to be always unique, which implies
398: that the transition is of second order, i.e., through a quantum critical line
399: in the $U$-$\mu$ plane.
400:
401:
402: Insight on the physical reason for the qualitative difference observed upon
403: particle or hole doping is obtained from the behavior of the observables
404: that measures the $d$-$p$ correlations. In Fig.\ref{fig4}
405: we show the behavior of magnetic correlation between the $d$ and
406: $p$ electrons
407: $\langle m^z_d m^z_p \rangle = \langle
408: (n_{d\uparrow} - n_{d\downarrow})
409: (n_{p\uparrow} - n_{p\downarrow}) \rangle $
410: across the metal-insulator transitions.
411: As the results show, the value of $\langle m^z_d m^z_p \rangle$ is
412: sizable on the hole doped metal, while becomes
413: negligible on the particle doped
414: side.
415: In the particle doped case $\langle m^z_d m^z_p \rangle$ is
416: negligible because the $p-$band is already full and the extra particles
417: mainly go to occupy $d-$sites.
418: Thus, the $p-$sites do not get an induced magnetic
419: moment and they cannot screen the magnetic moment of the local $d-$sites. Thus
420: the magnetic correlations develop directly among neighboring $d$ orbitals.
421: These correlations are of antiferromagnetic character due to the superexchange
422: mechanism and are analogous to those created between neighboring sites in the
423: Hubbard model case.
424: Thus we can understand that for particle doping the character
425: of the MIT in the Periodic Anderson and Hubbard models is in fact the same;
426: the $p-$sites merely allow the charge fluctuation (and thus the delocalization)
427: of the $d-$electrons, but they do not couple magnetically and do not screen the
428: local moments.
429:
430: In striking contrast, upon hole doping the scenario is quite different.
431: In this case, the system finds it is energetically
432: favorable to create holes in the $p-$band and magnetically bound the holes to
433: the local moment of the correlated $d-$site. This feature is
434: reminiscent of
435: the ``Zhang-Rice'' singlet formation \cite{zr} and leads to the emergence of a
436: quasiparticle peak at the Fermi energy as the singlets delocalize.
437:
438: In conclusion, we have investigated the doping driven metal-insulator transition
439: in the Periodic Anderson model in the Mott-Hubbard regime.
440: We found that the size of the Mott gap can be significantly
441: renormalized by hybridization effects. In addition, we found that while
442: both correlated metallic states at small doping show a
443: small quasiparticle peak at the Fermi energy, the
444: nature of the MIT is qualitatively different on each side.
445: In the particle doped side the quasiparticle peak is associated with a
446: Kondo-like resonance and the MIT shares the same qualitative nature
447: of the first order transition found in the one band Hubbard model.
448: In contrast, on the hole doped side, the quasiparticle peak
449: is associated with the formation of ``Zhang-Rice'' singlets
450: and the transition is second order.
451: Thus, our study demonstrates that, even in relatively simple situations,
452: the one band Hubbard model should not be automatically considered
453: the low energy effective model of more complicated
454: multi-orbital systems. The investigation of the physical nature of the
455: ``Zhang-Rice'' correlated metal is a very interesting problem open
456: for future investigations.
457: %This conclusion is of importance for the
458: %implementation of {\it ab initio} realistic band structure
459: %methods for strongly correlated electron systems.
460:
461: \begin{figure}%[!ht]
462: \centering
463: \includegraphics[width=8cm,clip]{fig4.eps}
464: \caption{$\langle m^z_d m^z_p \rangle$ as a function of $\mu$,
465: for $U=2$, $\Delta=1$, $t_{pd}=0.9$.
466: The results are obtained from ED with a cluster of 16 sites.}
467: \label{fig4}
468: \end{figure}
469:
470:
471: We acknowledge M. Gabay for useful discussions.
472:
473:
474: \begin{thebibliography}{99}
475:
476: \bibitem{mott}
477: N.F. Mott,
478: {\sl Metal-Insulator Transitions}
479: (Taylor \& Francis, London, 1974).
480:
481: \bibitem{ift}
482: For a review, see M. Imada, A. Fujimori and Y. Tokura,
483: {\sl Rev. Mod. Phys.} {\bf 70}, 1039 (1998).
484:
485: \bibitem{rmp}
486: For a review, see A. Georges, G. Kotliar, W. Krauth and M.J. Rozenberg,
487: {\sl Rev. Mod. Phys.} {\bf 68}, 13 (1996).
488:
489: \bibitem{mv}
490: W. Metzner and D. Vollhardt,
491: {\sl Phys. Rev. Lett.} {\bf 62}, 324 (1989).
492:
493: %\bibitem{uc2}
494: %X.Y. Zhang, M.J. Rozenberg and G. Kotliar,
495: %{\sl Phys. Rev. Lett.} {\bf 70}, 1666 (1993).
496:
497: \bibitem{ucs}
498: A. Georges and W. Krauth,
499: {\sl Phys. Rev. B} {\bf 48}, 7167 (1993);
500: M.J. Rozenberg, G. Kotliar and X.Y. Zhang,
501: {\sl Phys. Rev. B} {\bf 49}, 10181 (1994).
502:
503: \bibitem{prls}
504: M. J. Rozenberg, R. Chitra, and G. Kotliar,
505: {\sl Phys. Rev. Lett.} {\bf 83}, 3498 (1999);
506: G. Kotliar, E. Lange, and M. J. Rozenberg,
507: {\sl Phys. Rev. Lett.} {\bf 84}, 5180 (2000).
508:
509: \bibitem{limelette}
510: P. Limelette {\sl et al.},
511: {\sl Science} {\bf 302}, 89 (2003);
512: P. Limelette {\sl et al.},
513: {\sl Phys. Rev. Lett.} {\bf 91}, 016401 (2003).
514:
515: \bibitem{zsa}
516: J. Zaanen, G. A. Sawatzky and J. W. Allen,
517: {\sl Phys. Rev. Lett.} {\bf 55}, 418 (1985).
518:
519: \bibitem{phystoday}
520: G. Kotliar and D. Vollhardt,
521: {\sl Physics Today} {\bf 57}, 53 (2004).
522:
523: \bibitem{br}
524: W. F. Brinkman and T. M. Rice,
525: {\sl Phys. Rev. B} {\bf 2}, 4302 (1970).
526:
527: \bibitem{zr}
528: F. C. Zhang and T. M. Rice,
529: {\sl Phys. Rev. B} {\bf 37}, 3759 (1988).
530:
531: \bibitem{gabyRMP}
532: G. Kotliar {\sl et al.},
533: {\sl Rev. Mod. Phys.} {\bf 78}, 865 (2006).
534:
535: \bibitem{gks}
536: A. Georges, G. Kotliar, and Q. Si,
537: {\sl Int. J. Mod. Phys.} {\bf 6}, 705 (1992).
538:
539: \bibitem{mutou-ono}
540: T. Mutou, H. Takahashi and D. S. Hirashima,
541: {\sl J. Phys. Soc. Jpn.} {\bf 66}, 2781
542: (1997);
543: Y. Ono, R. Bulla and A.C. Hewson, {\sl Eur.
544: Phys. J. B} {\bf 19}, 375 (2001).
545:
546: %\bibitem{hf}
547: %J. Hirsch and R. Fye,
548: %{\sl Phys. Rev. Lett.} {\bf 56}, 2521 (1986).
549:
550: %\bibitem{ed}
551: %M. Caffarel and W. Krauth,
552: %{\sl Phys. Rev. Lett.} {\bf 72}, 1545 (1994);
553: %Q. Si, M. J. Rozenberg, G. Kotliar, and A. Ruckenstein,
554: %{\sl Phys. Rev. Lett.} {\bf 72}, 2761 (1994).
555:
556: %\bibitem{sahana}
557: %G. Kotliar, S. Murthy and M. J. Rozenberg,
558: %{\sl Phys. Rev. Lett.} {\bf 89}, 046401 (2002).
559:
560:
561:
562: \end{thebibliography}
563:
564: %\newpage
565:
566:
567: %
568: \end{document}
569:
570: