1: \documentclass[superscriptaddress,twocolumn,aps,pra,amsmath,floatfix]{revtex4}
2: \usepackage{graphicx}
3: \begin{document}
4:
5: \title{Surface vs. bulk Coulomb correlations in photoemission spectra
6: of perovskites}
7: % of SrVO$_3$ and CaVO$_3$}
8:
9: \author{A. Liebsch}
10: \affiliation{Institut f\"ur Festk\"orperforschung, Forschungszentrum
11: J\"ulich, 52425 J\"ulich, Germany}
12: \date{\today }
13:
14: \begin{abstract}
15: Recent photoemission spectra of the perovskite series
16: Sr$_x$Ca$_{1-x}$VO$_3$ revealed strong modifications associated with
17: surface contributions. To study the effect of Coulomb correlations
18: in the bulk and at the surface the quasi-particle spectra are evaluated
19: using the dynamical mean field theory. It is shown that as a result of
20: the reduced coordination number of surface atoms correlation effects
21: are stronger at the surface than in the bulk, in agreement with
22: experiment.\\
23:
24: \noindent
25: PACS numbers: 79.60.Bm, 71.27.+a, 71.20.be
26: \end{abstract}
27:
28: \maketitle
29:
30: Photoemission is a key spectroscopy for the investigation
31: of electronic properties of strongly correlated materials. Because
32: of its surface sensitivity there is growing concern to what
33: extent the data are representative of bulk properties and
34: whether they need to be corrected for surface effects influencing the
35: one- and many-electron properties. To study these questions Maiti
36: {\it et al.}\ \cite{maiti} and Sekiyama {\it et al.}\
37: \cite{sekiyama} performed
38: photoemission measurements on Sr$_x$Ca$_{1-x}$VO$_3$ using a
39: wide range of photon energies. As a result of the frequency dependent
40: escape depth of the photoelectron, the spectra reveal striking
41: variations evidently associated with surface modifications of the
42: electronic structure: emission from the coherent peak near the
43: Fermi level is reduced and correlation-induced satellites have larger
44: weight than in the bulk. The same trend is observed on
45: SrRuO$_3$\ \cite{fujioka}. Another example is Sr$_2$RuO$_4$ for which
46: earlier photoemission data appeared to be in conflict with de Haas-%
47: van Alphen measurements\ \cite{SrRuO1}. Recent experimental and theoretical
48: work proved that this discrepancy can be resolved by accounting for
49: the surface reconstruction of Sr$_2$RuO$_4$ which leads
50: to significant changes in photoemission spectra\ \cite{SrRuO2}.
51:
52: The understanding of surface effects in photoemission spectra of
53: transition metal oxides is important in order to distinguish them
54: from correlation phenomena in the bulk. Although the electronic
55: properties of strongly correlated materials are currently a field of
56: intense experimental and theoretical investigation\
57: \cite{imada,georges,held}
58: surface effects have so far received little attention\ \cite{potthoff}.
59: For this purpose photoemission studies of the series
60: Sr$_x$Ca$_{1-x}$VO$_3$\ \cite{maiti,sekiyama,fujimori,morikawa,inoue}
61: are very useful since the one-electron properties are relatively simple
62: and the ratio $U/W$ can be tuned systematically. Here, $U$ is the
63: on-site Coulomb energy and $W$ the width of the relevant bands near the
64: Fermi level. While SrVO$_3$ is considered to be a moderately correlated
65: metal with a clear quasi-particle peak near $E_F$, the slightly smaller
66: band width of CaVO$_3$ is believed to push this system closer
67: to the metal--insulator transition. For this reason CaVO$_3$ has become a
68: key system of investigation\ \cite{fukushima,shirakawa,inoue02}.
69: For instance, recent de Haas-van Alphen measurements
70: on CaVO$_3$\ \cite{inoue02} detected a cubic Fermi surface although
71: this system exhibits orthorombic distortions.
72:
73: In the present work we use the dynamical mean field theory (DMFT)
74: \ \cite{georges,vollhardt,pruschke} to evaluate the surface and bulk
75: quasi-particle spectra of SrVO$_3$. We show that because of the
76: planar electronic structure of perovskite systems and the concomitant
77: narrowing of the surface local density of states correlation
78: effects at the surface are more pronounced than in the bulk.
79: In the perovskites this enhancement is particularly strong since
80: the local Coulomb energy is not far from the critical value for a
81: metal-insulator transition. Such a
82: trend had been predicted by Potthoff and Nolting\ \cite{potthoff}
83: who studied the metal--insulator phase diagram for a semi-infinite
84: simple cubic $s$ band at half filling. Here we calculate the self-energy
85: for a multi-band system using a realistic local density of states
86: appropriate for SrVO$_3$ and find qualitative agreement with
87: photoemission data\ \cite{maiti,sekiyama}.
88:
89: Electronic structure calculations for SrVO$_3$ within the local density
90: approximation (LDA)\ \cite{takegahara} show that the conduction bands
91: near $E_F$ consist of degenerate $t_{2g}$ bands derived from V$^{4+}$
92: ($3d^1$) ions. The filled O 2p bands are separated from the $t_{2g}$
93: levels by a gap of about 1~eV, and the cubic crystal field of the V-O
94: octahedron shifts the $e_g$ bands above the $t_{2g}$ bands. The $t_{2g}$
95: bands can be represented via a tight-binding Hamiltonian with diagonal
96: elements
97: \,$h_{xy,xy}(k)= e_d + t_0 (c_x + c_y) + t_1 c_x c_y
98: + [t_2 + t_3(c_x + c_y) + t_4 c_x c_y] c_z$,
99: where $c_i = 2\cos(k_ia),\, i=x,y,z$ and $a$ is the lattice constant.
100: Cyclic permutations yield $h_{xz,xz}(k)$ and $h_{yz,yz}(k)$.
101: The $t_i$ denote effective hopping integrals representing the V-O-V
102: hybridization, where $t_{0,2}$, $t_{1,3}$, and $t_4$ specify the
103: interaction between first, second and third neighbors, respectively.
104: For symmetry reasons off-diagonal elements arise only between second
105: and third nearest neighbors and are of the form \,$h_{xy,xz}(k) =
106: - t'_1 s_y s_z - t'_2 c_x s_y s_z$, where $s_i = 2\sin(k_ia)$, i.e.,
107: they vanish at high-symmetry points. Since the $t'_{1,2}$ are very
108: small we neglect the off-diagonal elements so that the energy bands are
109: given by \,$\epsilon_i(k) = h_{i,i}(k)$, with \,$i=xy,xz,yz$.
110: The parameters $e_d$ and $t_0\cdots t_4$ can be found by fitting the
111: LDA energies at high-symmetry points of the Brillouin Zone.
112:
113: \begin{figure}
114: \begin{center}
115: \includegraphics[width=5cm,height=8cm,angle=-90]{Fig1a.ps}
116: \includegraphics[width=5cm,height=8cm,angle=-90]{Fig1b.ps}
117: \end{center}
118: \vskip-0.4cm
119: \caption{
120: (a) Tight-binding fit to LDA $t_{2g}$ bulk bands of SrVO$_3$ ($E_F=0$).
121: (b) Solid curve: isotropic bulk density of states;
122: dashed and dotted curves: local density of $d_{xz,yz}$ and $d_{xy}$
123: states in the first layer of SrVO$_3$.
124: }\end{figure}
125:
126:
127: Fig.~1\,(a) shows the $t_{2g}$ bulk bands of SrVO$_3$. According to the
128: predominantly planar electronic structure, the bulk density of states
129: $\rho_b(\omega)$ exhibits the characteristic asymmetric peak related
130: to the van Hove singularity at the X point (see Fig.~1\,(b)). The overall
131: shape of the bulk density of states agrees with LDA calculations\
132: \cite{takegahara}. At the surface, the $t_{2g}$ degeneracy is lifted
133: since only the $d_{xy}$ band exhibits strong dispersion within
134: the surface plane (the $z$ axis defines the surface normal).
135: To evaluate the surface density of states we use a Green's function
136: formalism\ \cite{kalkstein} for semi-infinite tight-binding systems.
137: As electronic structure calculations for SrVO$_3$ surfaces are not
138: available, we assume the tight-binding parameters to coincide with
139: those in the bulk, except for a small surface potential to ensure
140: charge neutrality.
141:
142: Fig.~1\,(b) shows the local density of states $\rho_s(\omega)$
143: of the $d_{xz,yz}$ bands for the surface layer of SrVO$_3$. Its
144: spectral weight is reduced at low and high energies but enhanced at
145: intermediate energies. Thus, the effective width of $\rho_s(\omega)$
146: is smaller than that of $\rho_b(\omega)$ although their total widths
147: are identical. In contrast to the $d_{xz,yz}$ density, the density
148: of $d_{xy}$ states nearly coincides with the bulk density, reflecting
149: the planar nature of the $t_{2g}$ states. In the deeper layers
150: the local density of $d_{xz,yz}$ states approaches $\rho_b(\omega)$
151: rather quickly, the main effect consisting in an oscillatory distortion
152: of the spectral shape. This rapid convergence is a consequence of the
153: tight-binding character of the $t_{2g}$ bands. It would be desirable
154: to carry out self-consistent electronic structure calculations for
155: SrVO$_3$ surfaces since they should provide a more accurate density
156: of states. Nevertheless, the key effect discussed here, namely, the
157: preferential narrowing of the local density of $d_{xz,yz}$ states,
158: should hold quite generally.
159:
160: To interpret the experimental photoemission data we evaluate the
161: quasi-particle spectra by accounting for local Coulomb correlations.
162: According to the one-electron properties discussed above we are
163: dealing with a non-isotropic system where two narrow bands interact
164: with a wider band. In the first layer this difference in
165: effective band width is largest and it diminishes rapidly towards
166: the interior of the system. This situation is reminiscent of the layer
167: perovskite Sr$_2$RuO$_4$, which essentially consists of Ru sheets
168: containing two narrow $d_{xz,yz}$ bands interacting with a wide
169: intra-planar $d_{xy}$ band. The peculiar feature of this system is
170: the fact the on-site Coulomb energy lies between the single-particle
171: widths of the non-degenerate $t_{2g}$ bands: \,$W_{xz,yz} < U < W_{xy}$
172: \ \cite{liebsch}. In the present case the difference
173: between $d_{xz,yz}$ and $d_{xy}$ bands at the surface is less
174: pronounced so that $W_i<U$ for all three bands. Nevertheless, since
175: in the first layer the effective width of $d_{xz,yz}$ states is
176: reduced, the influence of Coulomb correlations on the surface bands
177: should be stronger than on the wider bulk bands.
178:
179: The key quantity characterizing the effect of Coulomb correlations
180: is the complex self-energy. Since we neglect the
181: weak hybridization between $t_{2g}$ bands, $\Sigma(\omega)$ is
182: diagonal in orbital space. Despite this simplification, a full
183: description of correlations near the surface would be exceedingly
184: complicated since, in principle, it would require a mixed momentum/real
185: space approach in order to handle the loss of translational symmetry
186: normal to the surface. This could be accomplished using a cluster
187: formalism where the semi-infinite system is represented via a
188: slab of finite thickness. Unfortunately, the planar character of
189: $t_{2g}$ orbitals leads to a slowly convergent local density of
190: $d_{xz,yz}$ states, with many spurious $1/\sqrt{\omega}$ van
191: Hove singularities stemming from the quasi-one-dimensional hopping
192: parallel to the surface. On the other hand, a cluster generalization
193: of the DMFT is feasible today only for very small cluster size.
194: We therefore ignore the momentum variation of the self-energy and
195: assume that, for a given layer, it depends only on the local density
196: of states within that layer\ \cite{potthoff}. The reduced
197: dimensionality at the surface enters the many-body calculation
198: via the layer-dependent density of states. In view of the local
199: Coulomb interaction this assumption seems justified for qualitative
200: purposes.
201:
202: Before discussing the quasi-particle spectra derived within the DMFT
203: it is useful to compare the frequency dependence of the self-energy
204: in the bulk and at the surface. For convenience we give here the
205: results obtained within non-self-consistent second-order perturbation
206: theory since this provides a clear identification of the various orbital
207: contributions to the diagonal elements $\Sigma_i(\omega)$\ \cite{liebsch}.
208: Fig.~2 illustrates the second-order self-energy for
209: $U=4.3$~eV and $J=0.7$~eV. Although the total widths of the bulk
210: and surface densities shown in Fig.~1 are the same, the
211: effective band narrowing at the surface yields a correspondingly
212: narrower distribution of Im\,$\Sigma_{xz,yz}(\omega)$ compared to
213: Im\,$\Sigma_b(\omega)$ for the isotropic bulk.
214: Because of the orbital mixing via the on-site Coulomb interaction
215: Im\,$\Sigma_{xy}(\omega)$ lies between these two distributions.
216: From the Kramers-Kronig relations it then follows that
217: Re\,$\Sigma_{xz,yz}(\omega)$ exhibits a more pronounced
218: minimum below the band bottom than Re\,$\Sigma_b(\omega)$.
219: Thus, as indicated by the line $\omega-\epsilon_k$ for a typical
220: band energy $\epsilon_k$, the surface satellite appears at a
221: lower value of $U$ than in the bulk. Analogous arguments hold for
222: the upper Hubbard peak. This qualitative picture is fully supported
223: by the more accurate DMFT results discussed below.
224:
225: \begin{figure}
226: \begin{center}
227: \includegraphics[width=5.0cm,height=8cm,angle=-90]{Fig2.ps}
228: \end{center}
229: \vskip-0.4cm
230: \caption{
231: Second-order self-energy of SrVO$_3$.
232: Solid curves: bulk $t_{2g}$ self-energy;
233: dashed (dotted) curves: $d_{xz,yz}$ ($d_{xy}$) self-energy
234: for surface layer. For clarity the real part is shifted down
235: by 3~eV. The straight line gives $\omega-\epsilon_k$ for
236: $\epsilon_k=-0.5$~eV.
237: }\end{figure}
238:
239:
240: In the DMFT the self-energy elements $\Sigma_i(\omega)$
241: are functionals of the bath Green's functions
242: \,${\cal G}_i^{-1} = G_i^{-1} + \Sigma_i$.
243: The quasi-particle densities of states are related to the local $G_i$
244: via \,$N_i(\omega)=-{\rm Im}\,G_i(\omega)/\pi$.
245: On-site Coulomb correlations are treated using the self-consistent
246: multi-band Quantum Monte Carlo method (for a review, see Ref.\
247: \cite{georges}). The temperature of the simulation is 12.5~meV with
248: 64 imaginary time slices and several runs using $10^5$ Monte Carlo
249: sweeps. The quasi-particle density of states is obtained via maximum
250: entropy reconstruction\ \cite{jarrell}.
251:
252: \begin{figure}
253: \begin{center}
254: \includegraphics[height=8cm,width=5.0cm,angle=-90]{Fig3a.ps}
255: \includegraphics[height=8cm,width=5.0cm,angle=-90]{Fig3b.ps}
256: \includegraphics[height=8cm,width=5.0cm,angle=-90]{Fig3c.ps}
257: \end{center}
258: \vskip-0.4cm
259: \caption{
260: Quasi-particle density of states of SrVO$_3$ derived from DMFT.
261: (a) bulk $t_{2g}$ states,
262: (b) surface $d_{xz,yz}$ states,
263: (c) surface $d_{xy}$ states.
264: Solid curves: $U=4.3$\,eV; dot-dashed curves: $U=4.0$\,eV;
265: dashed curves: bare densities.
266: }\end{figure}
267:
268: Fig.~3\,(a) shows the bulk quasi-particle density of states of
269: SrVO$_3$ for two Coulomb energies: $U=4.0$\,eV and
270: $4.3$\,eV. The exchange energy is $J=0.7$\,eV\ \cite{zaanen}.
271: These results show that in the bulk $U$ must be larger than 4\,eV
272: to obtain the satellite observed in photoemission spectra
273: \cite{maiti,sekiyama}. The peak near 2~eV above $E_F$ agrees with
274: inverse photoemission data\ \cite{morikawa}. Although for $U=4$\,eV
275: there is considerable correlation-induced band narrowing and an
276: emerging satellite shoulder, the increase of $U$ to $4.3$\,eV yields
277: an even narrower coherent feature near $E_F$, with the missing weight
278: shifted to the lower and upper Hubbard bands. Note that
279: $N_b(E_F)=\rho_b(E_F)$ which follows from the local approximation
280: implicit in the DMFT for isotropic systems\ \cite{pinning}. The most
281: recent photoemission data\ \cite{sekiyama} confirm this result. The
282: spectra shown in Fig.~3\,(a) qualitatively agree with DMFT-LDA results
283: for the $3d^1$ perovskite La$_x$Sr$_{1-x}$TiO$_3$\ \cite{nekrasov}
284: which exhibits a similar $t_{2g}$ bulk density of states.
285:
286: The surface quasi-particle spectra for SrVO$_3$ are shown in
287: Fig.~3\,(b) and (c). The lower Hubbard peak of the $d_{xz,yz}$ states
288: is clearly visible already for $U=4$\,eV because of the narrower local
289: density of states in the first layer. A larger $U$ shifts the
290: satellite to higher binding energies. The comparison with the spectra
291: shown in Fig.~3\,(a) demonstrates that correlation effects for a fixed
292: value of $U$ are stronger at the surface than in the bulk: The coherent
293: peak near $E_F$ is narrower at the surface and the incoherent satellite
294: feature is more pronounced than in the bulk, in agreement with
295: experiment\ \cite{maiti,sekiyama,fujioka}.
296:
297: The surface quasi-particle density of $d_{xy}$ states in Fig.~3\,(c)
298: is intermediate between $N_b(\omega)$ and $N_s(\omega)$ for
299: $d_{xz,yz}$. Although there is little single-electron hybridization
300: between $t_{2g}$ bands, the local Coulomb interaction mixes them so that
301: the $d_{xy}$ surface spectrum involves contributions arising from the
302: more strongly correlated $d_{xz,yz}$ states. Another consequence
303: of the anisotropic surface self-energy is the
304: correlation-induced charge transfer between subbands\ \cite{liebsch}.
305: Here, we find that 0.06 electrons are shifted from the $d_{xy}$ to
306: the $d_{xz,yz}$ bands. Also, the quasi-particle partial densities of
307: states at $E_F$ do not need to coincide with the bare partial densities.
308: The coupling between narrow and wide bands is a genuine multi-band effect
309: and underlines the fact that single-particle bands in the presence of
310: local Coulomb interactions cannot be considered independently. This
311: issue will be considered in more detail elsewhere\ \cite{future}.
312:
313: Note that the many-body reduction of the quasi-particle band width is
314: much larger than the surface-induced one-electron band narrowing.
315: On the other hand, since with $U/W\approx2$ the on-site Coulomb
316: energy is not far from the critical value for a metal-insulator
317: transition \cite{georges}, the band narrowing substantially
318: enhances the influence of correlations at the surface.
319:
320: It would be interesting to perform angle-resolved photoemission
321: measurements to determine the correlation-induced band narrowing
322: of the SrVO$_3$ $t_{2g}$ bands. For instance the energy at $\bar\Gamma$
323: should be only a few tenths of an eV below $E_F$ instead of 1\,eV
324: as predicted by the LDA. Accordingly, the true $t_{2g}$ bands should be
325: considerably flatter than the single-particle bands. Also, measurements
326: using polarized light could help to separate correlation effects in the
327: $d_{xy}$ and $d_{xz,yz}$ bands.
328:
329: We finally comment on the difference between CaVO$_3$ and SrVO$_3$.
330: According to LDA band structure calculations for CaRuO$_3$ and
331: SrRuO$_3$\ \cite{mazin}, orthorombic distortions cause a slight
332: narrowing of the $t_{2g}$ bands and a strong broadening of the van
333: Hove singularity above $E_F$. Assuming the same on-site Coulomb energy
334: for the V$^{4+}$ ions in CaVO$_3$ and SrVO$_3$, a smaller bulk band
335: width of CaVO$_3$ implies a transfer of spectral weight from the
336: coherent to the incoherent peak. At present, the available data\
337: \cite{maiti,sekiyama,fujimori,morikawa,inoue}
338: do not yet fully agree on the magnitude of this effect.
339: At the surface, such a transfer of spectral weight should be more
340: pronounced because of the band narrowing discussed in the present paper.
341: This effect is indeed observed in the photoemission spectra\
342: \cite{maiti,sekiyama}. For a more detailed analysis it would be
343: important to carry out electronic structure calculations for
344: CaVO$_3$ in order to study orthorombic distortions
345: and possible reconstructions at the surface.
346:
347: In summary, we have performed DMFT quasi-particle calculations for
348: SrVO$_3$ in the bulk and at the surface. As a result of the planar
349: nature of the $t_{2g}$ states and the surface narrowing of the local
350: density of states on-site correlation effects are more pronounced at
351: the surface than in the bulk, in agreement with photoemission data.
352: The two-dimensional character of $d$ states near
353: $E_F$ is one of the hallmarks of transition metal oxides. The
354: surface-induced enhancement of correlation effects discussed in the
355: present work should therefore be a phenomenon observable in many
356: materials.
357:
358: I thank A. Bringer for discussions and A.I. Lichtenstein
359: for the QMC-DMFT code. I also thank S. Biermann for comments and for
360: bringing Ref.~2 to my attention.
361: \begin{thebibliography}{99}
362: \bibitem{maiti}
363: K. Maiti, D.D. Sarma, M.J. Rozenberg, I.H. Inoue, H. Makino, O. Goto,
364: M. Pedio, and R. Cimino,
365: Europhys. Lett. {\bf 55}, 246 (2001).
366:
367: \bibitem{sekiyama}
368: A. Sekiyama, H. Fujiwara, S. Imada, H. Eisaki, S.I. Uchida, K. Takegahara,
369: H. Harima, Y. Saitoh, and S. Suga,
370: cond-mat/0206471.
371:
372: \bibitem{fujioka}
373: K. Fujioka {\it et al.}, Phys. Rev. Lett. {\bf 56}, 6380 (1997).
374:
375: \bibitem{SrRuO1}
376: A.P. Mackenzie {\it et al.}, Phys. Rev. Lett. {\bf 78}, 2271 (1997);
377: T. Yokoya {\it et al.}, Phys. Rev. Lett. {\bf 78}, 2272 (1997);
378: and references herein.
379:
380: \bibitem{SrRuO2}
381: K.M. Shen {\it et al.}, Phys. Rev. B {\bf 64}, 180502(R) (2001);
382: A. Liebsch, Phys. Rev. Lett. {\bf 87}, 239701 (2001);
383: R. Matzdorf {\it et al.}, Science {\bf 289}, 746 (2000);
384: Phys. Rev. B {\bf 65}, 085404 (2002).
385:
386: \bibitem{imada}
387: M. Imada, A. Fujimori, and Y. Tokura, Rev. Mod. Phys. {\bf 70}, 1039 (1998).
388:
389: \bibitem{georges}
390: A. Georges, G. Kotliar, W. Krauth, and M.J. Rozenberg,
391: Rev. Mod. Phys. {\bf 68}, 13 (1996).
392:
393: \bibitem{held}
394: K. Held {\it et al.}, in {\it Quantum Simulations of Complex Many-Body
395: Systems: From Theory to Algorithms}, eds. J. Grotendorst, D. Marks,
396: and A. Muramatsu, NIC Series {\bf 10}, 175 (2002).
397:
398: \bibitem{potthoff}
399: M. Potthoff and W. Nolting,
400: Phys. Rev. B {\bf 60}, 7834 (1999);
401: Z. Phys. B {\bf 104}, 265 (1997).
402:
403: \bibitem{fujimori}
404: A. Fujimori {\it et al.},
405: Phys. Rev. Lett. {\bf 69}, 1796 (1992).
406:
407: \vskip2cm
408:
409: \bibitem{morikawa}
410: K. Morikawa {\it et al.},
411: Phys. Rev. B {\bf 52}, 13711 (1995).
412:
413: \bibitem{inoue}
414: I.H. Inoue {\it et al.}, Phys. Rev. Lett. {\bf 74}, 2539 (1995);
415: Phys. Rev. B {\bf 58}, 4372 (1998).
416:
417: \bibitem{fukushima}
418: A. Fukushima {\it et al.},
419: J. Phys. Soc. Jpn. {\bf 63}, 409 (1994).
420:
421: \bibitem{shirakawa}
422: N. Shirakawa {\it et al.},
423: J. Phys. Soc. Jpn. {\bf 64}, 4824 (1995).
424:
425: \bibitem{inoue02}
426: I.H. Inoue, C. Bergemann, I. Hase, and S.R. Julian,
427: Phys. Rev. Lett. {\bf 88}, 236403 (2002).
428:
429: \bibitem{vollhardt}
430: D. Vollhardt, in {\it Correlated Electron Systems}, edited by V.J. Emery
431: (World Scientific, Singapore, 1993), p. 57.
432:
433: \bibitem{pruschke}
434: Th. Pruschke, M. Jarrell, and J.K. Freericks,
435: Adv. in Phys. {\bf 44}, 187 (1995).
436:
437: \bibitem{takegahara}
438: K. Takegahara, J. Electron Spectrosc. Relat. Phenom. {\bf 66}, 303 (1994).
439:
440: \bibitem{kalkstein}
441: D. Kalkstein and P. Soven, Surf. Science {\bf 26}, 85 (1971).
442:
443: \bibitem{liebsch}
444: A. Liebsch and A. Lichtenstein, Phys. Rev. Lett. {\bf 84}, 1591 (2001).
445:
446:
447: \bibitem{jarrell}
448: M. Jarrell and J.E. Gubernatis, Phys. Rep. {\bf 269}, 133 (1996).
449:
450: \bibitem{zaanen}
451: J. Zaanen and G.A. Sawatzky, J. Solid State Chem. {\bf 88}, 8 (1990);
452: T. Mizokawa and A. Fujimori, Phys. Rev. B {\bf 54}, 5368 (1996).
453:
454: %\bibitem{kanamori}
455: %J. Kanamori, Progr. Theor. Phys. {\bf 30}, 275 (1963).
456:
457: \bibitem{pinning}
458: E. M\"uller-Hartmann, Z. Phys. {\bf 74}, 507 (1989).
459: % {\bf 76}, 211 (1989)
460:
461: \bibitem{nekrasov}
462: I.A. Nekrasov {\it et al.}, Euro. Phys. J. B {\bf 18}, 55 (2000).
463:
464: \bibitem{mazin}
465: I.I. Mazin and D.J. Singh,
466: Phys. Rev. B {\bf 56}, 2556 (1997).
467:
468: \bibitem{future}
469: A. Liebsch and S. Biermann, (to be published).
470:
471: \end{thebibliography}
472: \end{document}
473: